San Luis & Delta-Mendota Water Authority v. Salazar

*860

MEMORANDUM DECISION RE CROSS MOTIONS FOR SUMMARY JUDGMENT (DOCS. 548, 549, 550, 658, & 661)

OLIVER W. WANGER, District Judge.

*861

TABLE OF CONTENTS

I. INTRODUCTION........................................................863

II. PROCEDURAL HISTORY................................................863

III. STATUS OF THE SPECIES..............................................866

IV. SUMMARY OF MOTION.................................................867

A. Plaintiffs’ Motion.....................................................867

B. DWR’s Motion.......................................................867

Y. STANDARD OF DECISION..............................................868

VI. BASIC LEGAL FRAMEWORK ...........................................868

A. Review under the APA................................................868

(1) Record Review ...................................................868

(2) Deference to Agency Expertise.....................................869

(3) General Obligations Under the ESA.................................870

(4) Best Available Science.............................................870

(5) Best Available Science Standards and the Application of Analytical/Statistical Methodologies ...............................873

VII. ANALYSIS..............................................................874

A. Challenges to the Effects Analysis & Related Challenges to the RPA Actions............................................................874

(1) Legal Requirements for a Project Effects Analysis....................874

(2) Best Available Science Challenges to the Effects Analysis and Related Challenges to the Justification Provided for the RPA Actions........................................................875

a. The BiOp’s General Conclusion that Entrainment by Project Operations Adversely Affects Smelt Survival & Recovery is Supported by the Record.....................................876

b. Population Level Analysis/Life-Cycle Modeling....................881

e. FWS’ Use of Raw Salvage Numbers .............................885

(1) Federal Defendants’ Argument that the Flow Prescriptions in Actions 1 and 2 are Otherwise Justified...................890

(2) Use of Raw Salvage Analyses in Justification for Action 3.....895

d. FWS’s Comparison of CALSIM II Data to DAYFLOW Data........896

(1) Was FWS’s Decision to Compare Calsim II to Dayflow Model Runs a Violation of the Best Available Science Requirement? ...........................................903

(2) Does the Use of Dayflow to Represent the Baseline in the Project Effects Analysis Improperly Attribute Past Effects to the Projects?...................................909

(3) Use of Comparisons Between CALSIM and DAYFLOW Model Outputs to Justify Imposition of Component 3 (Action 4), the Fall X2 Action..............................910

(3) Other Challenges to the Fall X2 Action ..............................913

a. Plaintiffs’ Argument that Action 4 is an “Untested Hypothesis.”.....913

b. FWS’ Reliance on the Feyrer Papers.............................913

c. Do the Studies Cited in the BiOp Support FWS’s Conclusion that Fall X2 Determines the Extent of Suitable Smelt Habitat?....................................................915

(1) Feyrer (2007)..............................................915

(2) The Feyrer (2008) Paper....................................917

(3) The Bennett (2005) Article..................................918

d. Does the Best Available Science Support the Assumption that X2 Is a Surrogate for Smelt Habitat?...........................918

a. Are Delta Smelt Habitat Limited? ...............................919

*862

b. FWS’ Use of a Linear Model Instead of a Multiplicative Stock-Recruit Model...............................................920

c. DWR’s Challenge to the BiOp’s Choice of X2 Location..............922

(4) Challenges to Turbidity Trigger.....................................923

(5) Challenges to the Incidental Take Limit/Selective Use of Data..........924

a. FWS’s Exclusion of Certain Data Points When Analyzing Entrainment................................................924

b. FWS’s Use of Data to Examine the Relationship Between OMR Flows and Salvage and Exclusion of that Data from the Incidental Take Limit Analysis................................925

e. DWR’s Additional Challenges the ITS............................928

(6) Challenges to the BiOp’s Analysis of the Hydrodynamic Effects of the Projects....................................................929

a. Project Operations as a Driver of Hydrodynamic Conditions in the Delta...................................................930

b. Treatment of Other Stressors ...................................932

(1) Predation Analysis.........................................934

(2) Aquatic Macrophytes.......................................934

(3) Microcystis................................................936

(7) Indirect Effects Analysis...........................................936

a. Effect of Project Operations on Delta Smelt Food Supplies..........937

b. Pollution and Contaminants.....................................940

(8) Critical Habitat as Independent Basis for RPA........................943

a. Identification of a Threshold For Adverse Modification/ Explanation of How Any Alleged Alteration To Critical Habitat Would Exceed that Threshold..........................944

b. Reliance On Assumptions Of Indirect Effects Without Providing Evidence That These Indirect Effects Are Reasonably Certain To Occur............................................946

c. Reliance on Analysis Of Entrainment and X2 in Support of the Adverse Modification Determination............................947

(9) Discretionary v. Nondiseretionary Actions............................947

B. Application of the RPA Regulations.....................................948

(1) FWS Did Not Explicitly Analyze Any of the Four Factors in the BiOp..........................................................948

(2) Compliance with § 402.02 ..........................................949

a. Jeopardy Factor (Fourth Factor)................................949

b. Non-Jeopardy Factors (Factors One Through Three) ..............949

c. There is no Procedural Requirement that FWS Accept, Consider, and/or Address Comments Regarding the BiOp or its RPA.....957

C. Stewart

&

Jasper Orchards’ Argument Re: Reasonable and Prudent Measures..........................................................957

D. Stewart & Jasper, et al.’s, Argument that FWS Illegally Arrogated Authority to Itself Over Bureau of Reclamation and California Department of Water Resources Operations ...........................958

E. Information Quality Act Claim.........................................959

(1) Legal Framework of the IQA.......................................959

(2) Right to Judicial Review Under the APA.............................960

a. APA § 701(a)(2)’s Exception for Agency Action “Committed to Agency Discretion by Law” Bars Judicial Review in this Case.....961

(3) To the Extent FFA Bases Any of its Claims against Reclamation on the ESA, Such Claims are Subject to the ESA’s Pre-Filing Requirements ..................................................964

F. Renewed Claim That FWS Violated NEPA..............................964

G. Reclamation’s Liability under the ESA..................................966

VIII. CONCLUSION 967

*863

I.

INTRODUCTION

These consolidated cases arise out of the continuing war over protection of the delta smelt

(Hypomesus

transpacificus), an ESA-threatened species, and associated impacts to the water supply for more than half of the State of California. Plaintiffs, San Luis & Delta Mendota Water Authority (“SLDMWD”) and Westlands Water District, Metropolitan Water District of Southern California, State Water Contractors (“SWC”), Coalition for a Sustainable Delta and Kern County Water Agency, Stewart & Jasper Orchards, Arroyo Farms, LLC, and King Pistacho Grove, and Family Farm Alliance, move for summary judgment on their numerous remaining claims against the United States Fish and Wildlife Service’s (“FWS”) December 15, 2008 Biological Opinion addressing the impacts of the coordinated operations of the federal Central Valley Project (“CVP”) and State Water Project (“SWP”) on the threatened delta smelt

(Hypomesus tran

spacificus). Doc. 550. Plaintiff-in-Intervention, the California Department of Water Resources (“DWR”) filed a separate motion for summary judgment on narrower grounds. Docs. 548 & 549. Federal Defendants, the United States Department of the Interior, FWS, and the United States Bureau of Reclamation (“Reclamation”), and Defendant Intervenors, Natural Resources Defense Council and The Bay Institute, oppose and cross move for summary judgment on all remaining claims. Docs. 658 & 661. Plaintiffs and DWR replied. Docs. 697 & 695. The motion came on for hearing on July 8 & 9, 2010. After oral argument, the parties submitted supplemental briefing on a limited set of issues. Docs. 746-49.

II.

PROCEDURAL HISTORY

FWS’s 2005 biological opinion (“2005 Smelt BiOp”) found that the proposed coordinated operations of the SWP and CVP will have no adverse effect on the continued existence and recovery of the Delta Smelt and its critical habitat. The 2005 BiOp was remanded to FWS as arbitrary and capricious. Order,

NRDC v. Kempthorne,

506 F.Supp.2d 322 (E.D.Cal. 2007), Doc. 323. Following an extensive evidentiary hearing, the Court issued an interim remedial order and Findings of Fact and Conclusions of Law (“Findings”), which covered, among other things, the effects on delta smelt of negative flows in Old and Middle Rivers (“OMR”), two dis-tributary channels of the San Joaquin River.

See

Interim Remedial Order Following Summary Judgment and Evidentiary Hearing (“Int. Rem. Order”),

NRDC v. Kempthorne,

Doc. 560, 2007 WL 4462391 (Dec. 14, 2007); Findings re: Delta Smelt ESA Remand and Reconsultation (“Int. Rem. Findings”),

NRDC v. Kempthorne,

Doc. 561, 2007 WL 4462395 (Dec. 14, 2007).

1

Reclamation and DWR were ordered, among other things, to implement a winter “pulse flow” in OMR of no more negative than -2,000 cubic feet per second (“cfs”), and to “operate the CVP and SWP to achieve a daily average net upstream (reverse) flow in the OMR not to exceed 5,000 cfs on a seven-day running average” during a defined period in the spring. Int.

*864

Rem. Order at 5-7;

see also

Int. Rem. Findings at 15-20.

FWS issued a new delta smelt biological opinion on December 15, 2008 (“2008 Smelt BiOp” or “BiOp”).

See

Administrative Record (“AR”) at 00001-00411.

2

This BiOp concluded that proposed CVP and SWP operations are “likely to jeopardize the continued existence of’ the delta smelt and “adversely modify” its critical habitat. BiOp at 276-79. The BiOp includes a required Reasonable and Prudent Alternative (“RPA”) designed to allow the projects’ continued operations without causing jeopardy to the species or adverse modification to its critical habitat.

Id.

at 279-85. The RPA includes operational components designed to reduce entrainment of smelt during critical times of the year by controlling (limiting) water exports from the Delta by the Projects.

Id. at

279-85.

Component

1, to protect of the adult delta smelt life stage, consists of two Actions related to OMR flows.

•

Action 1,

to protect upmigrating delta smelt, is triggered during low and high entrainment risk periods based on physical and biological monitoring. Action 1 requires OMR flows to be no more negative than -2,000 cfs on a 14-day average and no more negative than -2,500 cfs for a 5-day running average.

Id. at

280-82, 329-51.

• Action 2,

to protect adult delta smelt that have migrated upstream and are present in the Delta prior to spawning. Action 2 is triggered immediately after Action 1 concludes or if recommended by the Smelt Working Group (“SWG”). Flows under Action 2 can be set within a range from -5,000 to -1,250 cfs, depending on a complex set of biological and environmental parameters.

Id. at

281-82, 352-56.

Component 2

(Action 3), to protect larval and juvenile delta smelt, requires OMR flows to be kept between -1,250 and -5,000 cfs, after Component 1 is completed, when Delta water temperatures reach 12° Celcius (“C”), or when a spent female smelt is detected in trawls or at salvage

3

facilities.

Id. at

282, 357-58. Component 2 continues until June 30 or when the Clifton Court Forebay water temperature reaches 25 ° C.

Id. at

282, 368.

Component 3

(Action 4), to improve habitat for delta smelt growth and rearing, requires sufficient Delta outflow to maintain average mixing point locations of Delta outflow and estuarine water inflow (“X2”

4

) from September to December, depending on water year type, in accordance with a specifically described “adaptive management process” overseen by FWS.

Id. at

282-83, 369.

5

*865

Component I

(Action 6) (Habitat Restoration), requires DWR to create or restore 8,000 acres of intertidal and subtidal habitat in the Delta and Suisun Marsh within 10 years.

Id. at

283-84, 379.

Component 5

(Monitoring and Reporting), requires Reclamation and DWR to gather and report information to ensure proper implementation of the RPA actions, achievement of physical results, and evaluation of the effectiveness of the actions on the targeted life stages of delta smelt, so that the actions can be refined, if needed.

Id. at

284-85, 328, 375.

The first of the six consolidated challenges to the BiOp was filed on March 3, 2009. Doc. 1. Plaintiffs moved for a preliminary injunction on April 24, 2009 to prevent Reclamation from implementing Component 2 of the RPA, alleging that FWS violated the National Environmental Policy Act (“NEPA”) and the ESA.

See

Doc. 31.

On May 22, 2009, the Court granted that motion in part, finding that Plaintiffs were likely to succeed on the merits of their NEPA claim and requiring FWS to make specific written findings to justify OMR flow restrictions.

See

Doc. 84;

see also

Doc. 94, Findings re Mot. for Prelim. Inj. (May 29, 2009). Defendants complied with that Order, submitting weekly notices of FWS’s OMR flow decisions.

See, e.g.,

Doc. Ill, Notice of OMR Flow Decision (June 11, 2009). The Court’s May 2009 preliminary injunction ruling was not based on Plaintiffs’ ESA claims. Doc. 94 at 43.

Plaintiffs amended their Complaint, joined and added claims against Reclamation,

see

Doc. 292, and moved for summary judgment on their NEPA claim,

see

Doc. 245. A November 13, 2009, 686 F.Supp.2d 1026 (E.D.Cal.2009) ruling granted summary adjudication in part, based on Reclamation’s failure to prepare an environmental impact statement before provisionally accepting and implementing the BiOp and its RPA Actions. Doc. 399.

Summary judgment for Defendants was granted on: (1) Stewart and Jasper Orchards’ Commerce Clause claim that the ESA did not apply to protect delta smelt, a purely intra-state species, Doc. 339; and (2) claims that the BiOp violated regulations governing formulation of the RPA by not including required information in the BiOp text, Doc. 354.

Plaintiffs then filed three temporary restraining order motions over a six week period—all of which were denied.

See

Docs. 555 & 583;

see also

3/16/10 Hrg. Tr. at 86-88. Plaintiffs next sought a preliminary injunction against implementation of RPA Component 3. An evidentiary hearing was held from April 2, 2010 through April 7, 2010. Docs. 644, 652-54. Findings Re Plaintiffs’ Request for Preliminary Injunction issued May 27, 2010 (“PI Decision”), 717 F.Supp.2d 1021 (E.D.Cal.2010). Doc. 704. The PI Decision confirmed Plaintiffs had succeeded on their NEPA claim and found Plaintiffs were likely to succeed on the merits of their ESA claim:

Although the premise underlying Component 2—that the species may be jeopardized by increased negative flows occasioned by export pumping—has record support, FWS has failed to adequately justify by generally recognized scientific principles the precise flow prescriptions imposed by Component 2. The exact restrictions imposed, which are inflicting material harm to humans and the human environment, are not supported by the record, making it impossible to determine whether RPA Component 2 [is] overly protective. Judicial deference is not owed to arbitrary, capricious, and scientifically unreasonable agency action.

*866

Id.

at 1070 . Plaintiffs presented evidence under NEPA on the balance of the hardships that social dislocation, unemployment, and other threats to human health and safety were caused by interdiction of Plaintiffs’ water supply.

See id.

at 1070 . Countervailing irreparable harm was found, because “the species and its critical habitat[] are entitled to protection under the ESA.”

Id.

at 1070 . Acknowledging the existence of legal and equitable grounds for injunctive relief, further evidence was requested on the “status of the species to assure that altered operations will not deepen jeopardy to the affected species or otherwise violate other laws.”

Id.

at 1071 . Specifically, to establish “that Plaintiffs’ proposed remedy of a flat -5,600 efs ceiling on negative OMR flows will not jeopardize the continued existence of the species and/or adversely modify its critical habitat.”

Id.

A May 28, 2010 status conference sought to determine whether a mutually-agreeable interim operational plan could be implemented. Doc. 706. On June 22, 2010, the parties stipulated to a joint operational plan to maintain OMR flows so as not to be more negative than -5,000 cfs, unless certain, defined salvage triggers required a further reduction in OMR flows. Doc. 724.

After these dispositive motions were filed, the National Academy of Sciences, completed a comprehensive review of the BiOp, and concluded that the BiOp and the RPA Actions were “scientifically justified.”

See

National Academy of Sciences, National Research Council, A Scientific Assessment of Alternatives for Reducing Water Management Effects on Threatened and Endangered Fishes in California’s Bay Delta at 3. Doc. 635. This post-decisional document is not part of the Administrative Record (“AR”) and no legal justification exists to supplement the AR to include it.

Additionally, a scientific peer review panel was convened by the private consulting firm, Post Buckley Shuh and Jernigan (“PBS

&

J”), at the request of Plaintiff Family Farm Alliance (“FFA”) in connection with FFA’s administrative petition under the Information Quality Act (“IQA”).

See Family Farm Alliance v. Salazar,

09-cv-1201 OWW-DLB (E.D.Cal.), Doc. 27, Ex. A. This document is part of the administrative record in the

Family Farm Alliance

IQA case, not the smelt AR. There is no basis to consider this document for non-IQA claims.

III.

STATUS OF THE SPECIES

The delta smelt was listed as a threatened species under the ESA on March 5, 1993. 58 Fed. Reg. 12,854 (March 5, 1993). Critical habitat was designated for the delta smelt on December 19, 1994. 59 Fed.Reg. 65,256 (Dec. 19, 1994). Once an abundant species in the Bay-Delta ecosystem as recently as thirty years ago, the delta smelt is now in imminent danger of extinction. PI Decision, Finding of Fact ¶ 10. All the evidence shows a significant decline in smelt abundance since 2000, recently up to three orders of magnitude below historic lows.

Id.

The latest fall mid-water trawl (“FMWT”) abundance index for the species was 17, the lowest level ever recorded.

Id.

On April 7, 2010, FWS announced that reclassifying the delta smelt from a threatened to an endangered species was warranted, but precluded by higher priority listing actions. 75 Fed.Reg. 17,667 (Apr. 7, 2010). The direct mortality of delta smelt by entrainment at the CVP-SWP pumps, as well as the destruction and adverse modification of its habitat in the Delta caused by water exports, were important factors in this determination.

Id.

at 17,669, 17,671 (“The operation of State and Federal export facilities constitute a

*867

significant and ongoing threat to delta smelt through direct mortality by entrainment”). As a result of the “immediate and high magnitude threats” confronting the species, the delta smelt was assigned a listing priority number of 2.

6

Id.

at 17,675.

IV.

SUMMARY OF MOTION

A. Plaintiffs’Motion.

Plaintiffs’ motion advances the following grounds and contentions:

(1) FWS failed to rely on the “best available science” by making fundamental scientific errors in its analysis of the impacts of Project Operations on the species by:

(a) Relying on raw salvage numbers in quantitative impact analyses;

(b) Failing to conduct a life cycle analysis;

(c) Comparing the results of two entirely different, incompatible flow and salinity models; and

(d) Selectively excluding certain data for one purpose, but then unjustifiably using it for another;

(2) The BiOp’s Project Effects Analysis is arbitrary and capricious because FWS:

(a) Assumed that Project operations drive hydrological conditions in the Delta and did not explain or justify this attribution;

(b) Evaluated the impacts of other (i.e., non-Project) stressors erroneously and inconsistently; and

(c) Improperly characterized summer food supply suppression, invasive species, and pollution and contaminants as indirect effects of Project Operations;

(3) The BiOp is arbitrary and capricious because it does not distinguish between discretionary and nondiscretionary actions, improperly inflating the alleged effects of Project Operations;

(4) The BiOp’s RPA is unlawful because FWS did not conduct the specific analyses required by the ESA and FWS’ own RPA regulation, 50 C.F.R. § 402.02 , because neither the BiOp nor the AR demonstrate that FWS analyzed or applied the first three (of four) § 402.02 factors;

(5) FWS illegally arrogated to itself Project operating authority in derogation of Reclamation and DWR;

(6) FWS acted arbitrarily and capriciously by disregarding the Information Quality Act (“IQA”) when preparing and issuing the BiOp;

(7) FWS violated NEPA by not considering the environmental impacts of issuing the BiOp and RPA.

(8) Reclamation violated its legal duties by accepting FWS’ inherently flawed BiOp.

B.

DWR’s Motion.

DWR’s attacks three aspects of the BiOp:

(1) By relying on a comparison of CALSIM II model runs with what the BiOp terms “historic” data (which was actually generated by the Day-flow model), the BiOp’s analysis of the effects of the proposed action on smelt habitat does not yield meaningful information and violates the ESA’s best available science requirement. This analysis further violates the APA because FWS did not adequately articulate any rational con

*868

nection between the facts found based on these comparisons, and its conclusions regarding the Projects’ effects on the smelt.

(2) Component 3 of the RPA, also referred to in the BiOp as Action 4, is intended to mitigate the effects of the proposed action on smelt habitat, by requiring the Projects to maintain X2 in specified locations, depending on the type of water year. The BiOp, however, lacks sufficient explanation as to the basis for the specific prescriptions imposed by this Component, in violation of the APA. Moreover, to the extent that the record reveals that these prescriptions are based, even in part, on the methods used in the effects analysis, they violate the ESA’s “best available science” mandate.

(3) The Incidental Take Statement (“ITS”) is defective. First, its estimates are based on the average take from water years 2006 through 2008, which predicts the ITS will likely be exceeded in half of all years. Second, FWS erroneously misapplied its own data with the result that the BiOp claims that the ITS was only exceeded in five of the previous sixteen years, rather than accurately stating that it was exceeded in eleven of the sixteen years. Third, the ITS take estimate is based on a data sample that is too small to provide a reasonable prediction of take under the RPA. These defects violate the ESA’s “best available science” requirement, the ESA’s ITS requirements, and the APA.

V.

STANDARD OF DECISION

Summary judgment is appropriate when the pleadings and the record demonstrate that “there is no genuine dispute as to any material fact and that the moving party is entitled to judgment as a matter of law.” Fed.R.Civ.P. 56(c). The claims in this case involve FWS’s issuance of a biological opinion, which is a final agency action subject to judicial review under the APA, 5 U.S.C. § 702 .

Nat'l Wildlife Fed’n v. Nat’l Marine Fisheries Serv.,

524 F.3d 917, 925 (9th Cir.2008)

(“NWF v. NMFS II”).

A court conducting judicial review under the APA may not resolve factual questions, but instead determines “whether or not as a matter of law the evidence in the administrative record permitted the agency to make the decision it did.”

Sierra Club v. Mainella,

459 F.Supp.2d 76, 90 (D.D.C.2006) (quoting

Occidental Eng’g Co. v. INS,

753 F.2d 766, 769 (9th Cir.1985)). “[I]n a case involving review of a final agency action under the [APA] ... the standard set forth in Rule 56(c) does not apply because of the limited role of a court in reviewing the administrative record.”

Id.

at 89. In this context, summary judgment becomes the “mechanism for deciding, as a matter of law, whether the agency action is supported by the administrative record and otherwise consistent with the APA standard of review.”

Id.

at 90.

VI.

BASIC LEGAL FRAMEWORK

A.

Review under the APA.

Administrative Procedure Act (“APA”) invalidation of a biological opinion requires Plaintiffs to prove that FWS’s action was “arbitrary, capricious, an abuse of discretion, or otherwise not in accordance with law.” 5 U.S.C. § 706 (2)(A).

(1)

Record Review.

APA review of a biological opinion is “based upon the evidence contained in the administrative record.”

Arizona Cattle Growers’ Ass’n v. FWS,

273 F.3d 1229 ,

*869

1245 (9th Cir.2001). Judicial review under the APA must focus on the administrative record already in existence, not some new record made initially in a reviewing court. Parties may not use “post-decision information as a new rationalization either for sustaining or attacking the agency’s decision.”

Ass’n of Pac. Fisheries v. EPA,

615 F.2d 794 , 811-12 (9th Cir.1980). Exceptions to administrative record review for technical information or expert explanation make such evidence admissible only for limited purposes, and those exceptions are narrowly construed and applied.

Lands Council v. Powell,

395 F.3d 1019 , 1030 (9th Cir.2005).

Here, as evidentiary rulings explained,

see, e.g.,

Docs. 387, 392 (10/19/09 Hrg. Tr.), 406, 407, 462, 740 (7/8/10 Hrg.), 750, expert testimony has been considered only for explanation of technical terms and complex scientific subject matter beyond the Court’s knowledge; and to understand the agency’s explanations, or lack thereof, and the parties’ arguments.

(2)

Deference to Agency Expertise.

A Court must defer to the agency on matters within the agency’s expertise, unless the agency completely failed to address some factor, consideration of which was essential to making an informed decision.

Nat’l Wildlife Fed’n v. Nat’l Marine Fisheries Serv.,

422 F.3d 782, 798 (9th Cir.2005)

(“NWF v. NMFS

I”). A court “may not substitute its judgment for that of the agency concerning the wisdom or prudence of the agency’s action.”

River Runners for Wilderness v. Martin,

593 F.3d 1064, 1070 (9th Cir.2010):

In conducting an APA review, the court must determine whether the agency’s decision is “founded on a rational connection between the facts found and the choices made ... and whether [the agency] has committed a clear error of judgment.”

Ariz. Cattle Growers’ Ass’n v. U.S. Fish & Wildlife,

273 F.3d 1229, 1243 (9th Cir.2001). “The [agency’s] action ... need be only a reasonable, not the best or most reasonable, decision.”

Nat’l Wildlife Fed. v. Burford,

871 F.2d 849 , 855 (9th Cir.1989).

Id.

Although deferential, judicial review under the APA is designed to “ensure that the agency considered all of the relevant factors and that its decision contained no clear error of judgment.”

Arizona v. Thomas,

824 F.2d 745, 748 (9th Cir.1987) (internal citations omitted). “The deference accorded an agency’s scientific or technical expertise is not unlimited.”

Brower v. Evans,

257 F.3d 1058, 1067 (9th Cir.2001) (internal citations omitted).

[An agency’s decision is] arbitrary and capricious if [it] has relied on factors which Congress has not intended it to consider, entirely failed to consider an important aspect of the problem, offered an explanation for its decision that runs counter to the evidence before the agency, or is so implausible that it could not be ascribed to a difference in view or the product of agency expertise.

Motor Vehicle Mfrs. Ass’n of U.S. v. State Farm Mut. Auto. Ins. Co.,

463 U.S. 29, 43 , 103 S.Ct. 2856 , 77 L.Ed.2d 443 (1983);

see also Citizens to Preserve Overton Park, Inc. v. Volpe,

401 U.S. 402, 416 , 91 S.Ct. 814 , 28 L.Ed.2d 136 (1971) (reviewing court may overturn an agency’s action as arbitrary and capricious if the agency failed to consider relevant factors, failed to base its decision on those factors, and/or made a “clear error of judgment”),

overruled on other grounds by Califano v. Sanders,

430 U.S. 99, 105 , 97 S.Ct. 980 , 51 L.Ed.2d 192 (1977).

More generally, “[u]nder the APA ‘the agency must examine the relevant data and articulate a satisfactory explana

*870

tion for its action including a rational connection between the facts found and the choice made.’ ”

Humane Soc. of U.S. v. Locke,

626 F.3d 1040, 1048 (9th Cir.2010) (quoting

Motor Vehicle Mfrs. Ass’n,

463 U.S. at 43 , 103 S.Ct. 2856 ). “The reviewing court should not attempt itself to make up for an agency’s deficiencies: We may not supply a reasoned basis for the agency’s action that the agency itself has not given.”

Id.

(3)

General Obligations Under the ESA.

ESA Section 7(a)(2) prohibits agency action that is “likely to jeopardize the continued existence” of any endangered or threatened species or “result in the destruction or adverse modification” of its critical habitat. 16 U.S.C. § 1536 (a)(2).

To “jeopardize the continued existence of’ means “to engage in an action that reasonably would be expected, directly or indirectly, to reduce appreciably the likelihood of both the survival and recovery of a listed species in the wild by reducing the reproduction, numbers, or distribution of that species.” 50 C.F.R. § 402.02 ;

see also NWF v. NMFS II,

524 F.3d 917 (rejecting agency interpretation of 50 C.F.R. § 402.02 that in effect limited jeopardy analysis to survival and did not realistically evaluate recovery, thereby avoiding an interpretation that reads the provision “and recovery” entirely out of the text). An action is “jeopardizing” if it keeps recovery “far out of reach,” even if the species is able to cling to survival.

NWF v. NMFS II,

524 F.3d at 931 . “[A]n agency may not take action that will tip a species from a state of precarious survival into a state of likely extinction. Likewise, even where baseline conditions already jeopardize a species, an agency may not take action that deepens the jeopardy by causing additional harm.”

Id.

at 930 .

To satisfy this obligation, the federal agency undertaking the action (the “action agency”) must prepare a “biological assessment” that evaluates the action’s potential impacts on species and species’ habitat. 16 U.S.C. § 1536 (c); 50 C.F.R. § 402.12 (a). If the proposed action “is likely to adversely affect” a threatened or endangered species or adversely modify its designated critical habitat, the action agency must engage in “formal consultation” with FWS to obtain its biological opinion as to the impacts of the proposed action on the listed species.

See

16 U.S.C. § 1536 (a)(2), (b)(3);

see also

50 C.F.R. § 402.14 (a), (g). Once the consultation process has been completed, FWS must give the action agency a written biological opinion “setting forth [FWS’s] opinion, and a summary of the information on which the opinion is based, detailing how the agency action affects the species or its critical habitat.” 16 U.S.C. § 1536 (b)(3)(A);

see also

50 C.F.R. § 402.14 (h).

If FWS determines that jeopardy or destruction or adverse modification of critical habitat is likely, FWS “shall suggest those reasonable and prudent alternatives which [it] believes would not violate subsection (a)(2) of this section and can be taken by the Federal agency or applicant in implementing the agency action.” 16 U.S.C. § 1536 (b)(3)(A). “Following the issuance of a ‘jeopardy’ opinion, the agency must either terminate the action, implement the proposed alternative, or seek an exemption from the Cabinet-level Endangered Species Committee pursuant to 16 U.S.C. § 1536 (e).”

Nat’l Ass’n of Home Builders v. Defenders of Wildlife,

551 U.S. 644, 652 , 127 S.Ct. 2518 , 168 L.Ed.2d 467 (2007).

(4)

Best Available Science.

Under the ESA, an agency’s actions must be based on “the best scientific

*871

and commercial data available.” 16 U.S.C. § 1536 (a)(2); 50 C.F.R. § 402.14 (g)(8) (“In formulating its Biological Opinion, any reasonable and prudent alternatives, and any reasonable and prudent measures, the Service will use the best scientific and commercial data available.... ”). A failure by the agency to utilize the best available science is arbitrary and capricious.

See Pac. Coast Fed’n of Fishermen’s Assns. v. Gutierrez,

606 F.Supp.2d 1122, 1144 (E.D.Cal.2008).

“The obvious purpose of the [best available science requirement] is to ensure that the ESA not be implemented haphazardly, on the basis of speculation or surmise.”

Bennett v. Spear,

520 U.S. 154, 176 , 117 S.Ct. 1154 , 137 L.Ed.2d 281 (1997).

While this no doubt serves to advance the ESA’s overall goal of species preservation, we think it readily apparent that another objective [of the best available science requirement] (if not indeed the primary one) is to avoid needless economic dislocation produced by agency officials zealously but unintelligently pursuing their environmental objectives. That economic consequences are an explicit concern of the ESA is evidenced by § 1536(h), which provides exemption from § 1536(a)(2)’s no-jeopardy mandate where there are no reasonable and prudent alternatives to the agency action and the benefits of the agency action clearly outweigh the benefits of any alternatives. We believe the “best scientific and commercial data” provision is similarly intended, at least in part, to prevent uneconomic (because erroneous) jeopardy determinations.

Id.

at 176-77, 117 S.Ct. 1154 .

A decision about jeopardy must be made based on the best science available at the time of the decision; the agency cannot wait for or promise future studies.

See Ctr. for Biological Diversity v. Rumsfeld,

198 F.Supp.2d 1139, 1156 (D.Ariz.2002). The “best available science” mandate of the ESA sets a basic standard that “prohibits the [agency] from disregarding available scientific evidence that is in some way better than the evidence [it] relies on.”

Am. Wildlands v. Kempthorne,

530 F.3d 991, 998 (D.C.Cir. 2008) (citation omitted).

What constitutes the “best” available science implicates core agency judgment and expertise to which Congress requires the courts to defer; a court should be especially wary of overturning such a determination on review.

Baltimore Gas & Flee. Co. v. Natural Res. Defense Council,

462 U.S. 87, 103 , 103 S.Ct. 2246 , 76 L.Ed.2d 437 (1983) (a court must be “at its most deferential” when an agency is “making predictions within its area of special expertise, at the frontiers of science”). As explained in the

en banc

decision in

Lands Council,

537 F.3d at 993, courts may not “impose on the agency their own notion of which procedures are best or most likely to further some vague, undefined public good.” In particular, an agency’s “scientific methodology is owed substantial deference.”

Gifford Pinchot Task Force v. U.S. Fish & Wildlife Serv.,

378 F.3d 1059 , 1066 (9th Cir.2004).

“When specialists express conflicting views, an agency must have discretion to rely on the reasonable opinions of its own qualified experts even if, as an original matter, a court might find contrary views more persuasive.”

Lands Council,

537 F.3d at 1000 (quoting

Marsh v. Oregon Natural Res. Council,

490 U.S. 360, 378 , 109 S.Ct. 1851 , 104 L.Ed.2d 377 (1989)). Mere uncertainty, or the fact that evidence may be “weak,” is not fatal to an agency decision.

Greenpeace Action v. Franklin,

14 F.3d 1324, 1337 (9th Cir.1992) (upholding biological opinion, despite uncertainty about the effectiveness of management

*872

measures, because decision was based on a reasonable evaluation of all available data);

Natl Wildlife Fed’n v. Babbitt,

128 F.Supp.2d 1274, 1300 (E.D.Cal.2000) (holding that the “most reasonable” reading of the best scientific data available standard is that it “permits the [FWS] to take action based on imperfect data, so long as the data is the best available”). FWS “must utilize the ‘best scientific ... data available,’ not the best scientific data possible.”

Building Indus. Ass’n v. Norton,

247 F.3d 1241, 1246 (D.C.Cir.2001), cited with approval in

Kern County Farm Bureau v. Allen,

450 F.3d 1072, 1080-81 (9th Cir. 2006) (“Absent superior data occasional imperfections do not violate” the ESA best available data standard);

see also Defenders of Wildlife v. Babbitt,

958 F.Supp. 670, 680 (D.D.C.1997) (best available science standard does not require “conclusive evidence,” only that agency use best science available and not ignore contrary evidence).

The deference afforded under the best available science standard is not unlimited. For example,

Tucson Herpetological Society v. Salazar,

566 F.3d 870, 879 (9th Cir.2009), held that an agency may not rely on “ambiguous studies as evidence” to support findings made under the ESA. Because the studies did not lead to the conclusion reached by FWS, the Ninth Circuit held that these studies provided inadequate support in the administrative record for the determination made by FWS.

Id.; see also Rock Creek Alliance v. U.S. Fish & Wildlife Service,

390 F.Supp.2d 993, 1008 (D.Mont.2005) (rejecting FWS’s reliance on a disputed scientific report, which explicitly stated its analysis was not applicable to the small populations addressed in the challenged opinion). Alternatively, the presumption of agency expertise may be rebutted if the agency’s decisions, although based on scientific expertise, are not reasoned,

Greenpeace v. NMFS,

80 F.Supp.2d 1137, 1147 (W.D.Wash.2000), or if the agency disregards available scientific evidence better than the evidence on which it relies,

Kern County Farm Bureau,

450 F.3d at 1080 .

Courts routinely perform substantive reviews of record evidence to evaluate the agency’s treatment of best available science. The judicial review process is not one of blind acceptance.

See, e.g., Kern County,

450 F.3d at 1078-79 (thoroughly reviewing three post-comment studies and FWS’s treatment of those studies to determine whether they “provide[d] the sole, essential support for” or “merely supplemented” the data used to support a listing decision);

Home Builders Ass’n of N. Cal. v. U.S. Fish and Wildlife Serv.,

529 F.Supp.2d 1110, 1120 (N.D.Cal.2007) (examining substance of challenge to FWS’s determination that certain data should be disregarded);

Trout Unlimited v. Lohn,

645 F.Supp.2d 929 (D.Or.2007) (finding best available science standard had been violated after thorough examination of rationale for NMFS’s decision to withdraw its proposal to list Oregon Coast Coho salmon);

Oceana, Inc. v. Evans,

384 F.Supp.2d 203, 217-18 (D.D.C.2005) (carefully considering scientific underpinnings of challenge to FWS’s use of a particular model, including post decision evidence presented by an expert to help the court understand the complex model, applying one of several record review exceptions articulated in

Esch v. Yeutter,

876 F.2d 976, 991 (D.C.Cir.1989), which are similar to those articulated by the Ninth Circuit).

Courts are not required to defer to an agency conclusion that runs counter to that of other agencies or individuals with specialized expertise in a particular technical area.

See, e.g., Am. Tunaboat Ass’n v. Baldrige,

738 F.2d 1013, 1016-17 (9th Cir.1984) (NMFS’s decision under the

*873

Marine Mammal Protection Act was not supported by substantial evidence because agency ignored data that was product of “many years’ effort by trained research personnel”);

Sierra Club v. U.S. Army Corps of Eng’rs,

701 F.2d 1011 , 1030 (2d Cir.1983) (“court may properly be skeptical as to whether an EIS’s conclusions have a substantial basis in fact if the responsible agency has apparently ignored the conflicting views of other agencies having pertinent experience! ]”) (internal citations omitted). A court should “reject conelusory assertions of agency ‘expertise’ where the agency spurns unrebutted expert opinions without itself offering a credible alternative explanation.”

N. Spotted Owl v. Hodel,

716 F.Supp. 479, 483 (W.D.Wash.1988) (citing

Am. Tunaboat Ass’n,

738 F.2d at 1016 ).

In

Conner v. Burford,

848 F.2d 1441, 1453-54 (9th Cir.1988), the agency attempted to defend its biological opinions by arguing that there was a lack of sufficient information to perform additional analysis. In rejecting this defense, the Ninth Circuit held that “incomplete information ... does not excuse the failure to comply with the statutory requirement of a comprehensive biological opinion using the best information available,” and noted that FWS could have completed more analysis with the information that was available.

Id.

at 1454 .

In light of the ESA requirement that the agencies use the best scientific and commercial data available ... the FWS cannot ignore available biological info or fail to develop projections of ... activities which may indicate potential conflicts between development and the preservation of protected species. We hold that the FWS violated the ESA by failing to use the best information available to prepare comprehensive biological opinions.

Id.

(emphasis added).

(5)

Best Available Science Standards and the Application of Analytical/Statistical Methodologies.

The above-described standards apply with equal force to the use and interpretation of statistical methodologies. As the D.C. Circuit in

Appalachian Power Co. v. EPA,

135 F.3d 791 (D.C.Cir.1998), explained in reviewing a challenge to a decision of the Environmental Protection Agency (“EPA”) under the “arbitrary and capricious” standard of review:

Statistical analysis is perhaps the prime example of those areas of technical wilderness into which judicial expeditions are best limited to ascertaining the lay of the land. Although computer models are “a useful and often essential tool for performing the Herculean labors Congress imposed on EPA in the Clean Air Act,” [citation] their scientific nature does not easily lend itself to judicial review. Our consideration of EPA’s use of a regression analysis in this case must therefore comport with the deference traditionally given to an agency when reviewing a scientific analysis within its area of expertise without abdicating our duty to ensure that the application of this model was not arbitrary.

Id.

at 802 .

The model must fit the available data.

See Nat'l Wildlife Fed’n v. EPA,

286 F.3d 554, 565 (D.C.Cir.2002)

(“NWF v. EPA

”) (a court will only reject the choice of a model “when the model bears no rational relationship to the characteristics of the data to which it was applied”). For example,

Oceana,

384 F.Supp.2d at 220 , rejected a challenge to NMFS’s use of a particular analytical model that used data drawn from existing literature, even though ex

*874

perts “suggested that reliable take limits cannot be established without quantitative data gathered from ‘in-water’ surveys.” Although NMFS conceded “a thorough quantitative analysis based on empirical estimates of population size would be a superior way to analyze the impact [ ] on [the species],” it was undisputed that “given the paucity of information on sea turtles and the difficulties of using the data that does exist, ‘[a] different or more complex model [than that used by NMFS] was not available and could not even be constructed.’ ”

Id.

Likewise, “the fact that a given model has some imperfections does not prevent it from constituting the ‘best scientific information available.’ ”

Oceana v. Evans,

2005 WL 555416 , *16-*17 (D.D.C. Mar. 9, 2005) (citing 16 U.S.C. § 1851 (a)(2)) (approving NMFS’s use of a model despite known limitations, where it was the only model available and the agency supplemented its analysis with other sources to address areas where the model was unable to make accurate predictions).

VII.

ANALYSIS

A.

Challenges to the Effects Analysis & Related Challenges to the RPA Actions.

(1)

Legal Requirements for a Project Effects Analysis.

Under section 7(a)(2) of the ESA and the Joint Consultation Regulations, FWS must “[e]valuate the effects of the action and cumulative effects on the listed species or critical habitat.” 50 C.F.R. § 402.14 (g)(3). FWS must then “[f]ormulate its biological opinion as to whether the action, taken together with cumulative effects,

7

is likely to jeopardize the continued existence of listed species or result in the destruction or adverse modification of critical habitat.” § 402.14(g)(4). The effects of the action are defined as:

the direct and indirect effects of an action on the species or critical habitat, together with the effects of other activities that are interrelated or interdependent with that action, that will be added to the environmental baseline.

§ 402.02.

The environmental baseline includes: the past and present impacts of all Federal, State, or private actions and other human activities in the action area, the anticipated impacts of all proposed Federal projects in the action area that have already undergone formal or early section 7 consultation, and the impact of State or private actions which are contemporaneous with the consultation in process.

Id.

The baseline is described in FWS and NMFS’s Joint Consultation Handbook

8

as:

an analysis of the effects of past and ongoing human and natural factors leading to the current status of the species, its habitat (including designated critical habitat), and ecosystem, within the action area. The environmental baseline is a “snapshot” of a species’ health at a specified point in time. It does not include the effects of the action under review in the consultation.

Consultation Handbook 4-22.

Once the baseline, the “direct and indirect effects” of the action, and the “effects

*875

of other activities that are interrelated or interdependent with that action” are determined, 50 C.F.R. § 402.02 , FWS then is required to consider whether, in light of the environmental baseline, the effects of the action, taken together with cumulative effects, are likely to jeopardize the continued existence of the listed species, 50 C.F.R. § 402.14 (g).

[An] agency may not take action that will tip a species from a state of precarious survival into a state of likely extinction. Likewise, even where baseline conditions already jeopardize a species, an agency may not take action that deepens the jeopardy by causing additional harm.

.... [The agency must] appropriately consider the effects of its actions “within the context of other existing human activities that impact the listed species.”

ALCOA [v. Administrator, Bonneville Power Admin.],

175 F.3d [1156,] 1162 n. 6 [(9th Cir.1999) ] (citing 50 C.F.R. § 402 .02’s definition of the environmental baseline). This approach is consistent with our instruction ... that “[t]he proper baseline analysis is not the proportional share of responsibility the federal agency bears for the decline in the species, but what jeopardy might result from the agency’s proposed actions

in the present and future human and natural contexts.” [PCFFA v. U.S. Bureau of Reclamation],

426 F.3d [1082,] 1093 [ (9th Cir.2005) ] (emphasis added).

NWF v. NMFS II,

524 F.3d at 930 (emphasis in original).

To jeopardize means “to engage in an action that reasonably would be expected, directly or indirectly, to reduce appreciably the likelihood of both the survival and recovery of a listed species.” 50 C.F.R. § 402.02 . The Consultation Handbook further provides that to “appreciably diminish the value: [means] to considerably reduce the capability of designated [critical habitat].” Consultation Handbook at 4-36. A related case found:

interpretation of “appreciably” to mean any “perceptible” effect would lead to irrational results, making any agency action that had any effects on a listed species a “jeopardizing” action. This is not the law, as such an interpretation conflicts with other provisions of the ESA that permit incidental take of listed species.

PCFFA v. Gutierrez,

606 F.Supp.2d 1195, 1208 (E.D.Cal.2008) (citing 16 U.S.C. 1536(b)(4), 1539(1)(B)).

(2)

Best Available Science Challenges to the Effects Analysis and Related Challenges to the Justification Provided for the RPA Actions.

Plaintiffs argue that the project effects analysis is predicated upon scientific errors that render the BiOp and its conclusion that project operations jeopardize the delta smelt arbitrary, capricious and an abuse of discretion:

The Project Effects Analysis is the heart of the section 7 consultation process, providing the basis for FWS’ jeopardy and adverse modification determinations and for formulating the RPA. In this case, FWS began the Project Effects Analysis of the 2008 Smelt BiOp with a remarkable assumption: “The following analysis assumes that the proposed CVP/SWP operations affect delta smelt throughout the year either directly through entrainment or indirectly through influences on its food supply and habitat suitability.” BiOp at 203 (AR 000218.) This assumption plainly violates the “best available science” required by the ESA. The science, including the reports that FWS purports to rely on, shows that OMR flows and entrainment do not have any statistically

*876

significant effect on the delta smelt’s population growth rate. Restricting flows has no effect on the delta smelt population’s survival-such restrictions are a costly, but meaningless gesture. The same is true for [restrictions designed to control the position of] X2 [in the Fall].

Doc. 551 at 8.

Plaintiffs maintain that the best available science does not support FWS’ “assumption” that “CVP/SWP operations affect delta smelt throughout the year either directly through entrainment or indirectly through influences on its food supply and habitat suitability.” BiOp at 203. Plaintiffs maintain that the science demonstrates:

(a) OMR flows have no statistically significant effect on the delta smelt population growth rate;

(b) With respect to the adult population, only OMR flows more negative than - 6,100 cfs will correlate to an increase in entrainment;

9

(c) The location of Fall X2 does not determine the extent and quality of suitable smelt habitat-as with OMR flows, Fall X2 has no statistically significant effect on the population growth rate; and,

(d) The CVP/SWP projects do not indirectly govern abiotic and biotic factors in the Delta that affect delta smelt abundance.

Doc. 551 at 11. Plaintiffs also maintain that there is no scientific support for the BiOp’s assumption that the Projects control hydrodynamic conditions in the Delta, or for the BiOp’s classification of non-Project causes of harm as “indirect effects” of Project Operations.

Id.

a.

The BiOp’s General Conclusion that Entrainment by Project Operations Adversely Affects Smelt Survival & Recovery is Supported by the Record.

The magnitude of diversions at the CVP and SWP pumping facilities influences flows throughout the Delta, including in the Old and Middle Rivers (“OMR”). BiOp at 160. When the level of diversion at the pumps is high, Old and Middle Rivers may flow backwards (in the opposite direction than they would under natural hydrological conditions) and toward the CVP and SWP natural conditions (called “negative” flows).

Id.

Negative OMR flows draw delta smelt present in the central and south Delta toward the pumps, and high negative flows increase the risk that they will be entrained at the pumps.

Id.

at 163, 253 (Figure E-7).

Unlike larger fish species, entrainment is lethal for weak-swimming delta smelt.

Id.

at 145. Relying on estimates of proportional entrainment presented by Dr. Wim Kimmerer in a 2008 paper entitled “Losses of Sacramento River Chinook Salmon and Delta Smelt to Entrainment in Water Diversions in the Sacramento-San Joaquin Delta,” published in the journal, San Francisco Estuary & Watershed Science (“Kimmerer (2008)”), the BiOp concludes that “[t]otal annual entrainment of the delta smelt population (adults and their progeny combined) ranged from approximately 10 percent to 60 percent per year from 2002-2006.”

Id.

at 210. In years when low flows and high exports coincide with a spawning distribution of the delta smelt that includes the San Joaquin River, the loss of larval delta smelt due to entrainment can exceed 50% of the population.

Id.

at 164-65. Such losses do not occur every year, but FWS concluded the effect of these large larval loss events

*877

is “substantial when it does,” particularly in light of the fact that the delta smelt is an annual fish.

Id.

at 165. Even one year where its spawning occurs “within the footprint of entrainment by the pumps” can lead to “a [severe] reduction in that year’s production.”

Id.

The BiOp’s Effects Analysis concludes that Project pumping operations have a “sporadically significant” adverse effect on smelt abundance:

The population-level effects of delta smelt entrainment vary; delta smelt entrainment can best be characterized as a

sporadically significant

influence on population dynamics. Kimmerer (2008) estimated that annual entrainment of the delta smelt population (adults and their progeny combined) ranged from approximately 10 percent to 60 percent per year from 2002-2006. Major population declines during the early 1980s (Moyle et al. 1992) and during the recent POD years (Sommer et al. 2007) were both associated with hydrodynamic conditions that greatly increased delta smelt entrainment losses as indexed by numbers of fish salvaged. However, currently published analyses of long-term associations between delta smelt salvage and subsequent abundance do not support the hypothesis that entrainment is driving population dynamics year in and year out ( Bennett 2005 ; Manly and Chotkowski 2006; Kimmerer 2008).

BiOp at 210 (emphasis added). This passage was based in large part on Kimmerer (2008), which states:

Delta smelt may suffer substantial losses to export pumping both as prespawning adults and as larvae and early juveniles. In contrast to the situation for salmon, pre-salvage mortality has been constrained in the calculations for adult Delta smelt, and its effects eliminated from the calculations for larval/juvenile Delta smelt. Combining the results for both life stages, losses may be on the order of zero to 40 percent of the population throughout winter and spring. The estimates have large confidence limits, which could be reduced by additional sampling, particularly to estimate 0 in Equation 18. If there is interest in improving these estimates further, some attempts should be made to examine the assumptions not fully tested above, particularly those used in extrapolating larval abundance to hatch dates.

AR 018877.

Plaintiffs argue that the BiOp misinterprets and misapplies Kimmerer’s work. Dr. Bryan Manly, Plaintiffs’ expert in the fields of biostatistics and population survey design, addressed the BiOp’s statement that “delta smelt entrainment can best be characterized as a sporadically significant influence on population dynamics.” Manly Decl., Doc. 897, at ¶ 7. Manly opines that “[t]his statement is unclear and confusing,” and explains:

If the Service meant only that abundance at a point in time during a single year may vary depending upon entrainment, then Kimmerer’s estimates support that statement. But if, as appears more likely, the Service was relying upon Kimmerer’s estimates to support a conclusion that entrainment sometimes causes abundance to vary significantly later in the same year or in following years, then the statement in the BiOp has no scientific basis.

Id.

Kimmerer (2008) only estimated percentage losses of delta smelt within single year classes, and did not conclude that such losses reduce population abundance from one year to the next.

Id.

at ¶ 8. In fact, Kimmerer (2008) contains a number of disclaimers, including the caveat that

*878

“export effects” on smelt are small relative to other factors affecting survival:

Although the upper bound of [the 0-40% loss] range represents a substantial loss, the effect of this loss is complicated by subsequent variability in survival (Figure 17). If this variability is uncorrelated with entrainment losses, then these losses will contribute little to the variability in fall abundance index. The simplest way to evaluate this is by regression of fall midwater trawl index on winter-spring export flow, but this relationship is contaminated by the downward step change in abundance in approximately 1981-1982, together with the long-term upward trend in export flow (mainly up to the mid-1970s,

see

Kimmerer 2004). Including this step in a regression model eliminates the effect of export flow on the fall midwater trawl index (coefficient = -1.5 ± 2.4, 95% CL, 36 df). It seems unlikely that the downward step change was due to the earlier increase in export flow; furthermore, despite substantial variability in export flow in years since 1982, no effect of export flow on subsequent mid-water trawl abundance is evident.

This is not to dismiss the rather large proportional losses of delta smelt that occur in some years; rather, it suggests that these losses have effects that are episodic and that therefore their effects should be calculated rather than inferred from correlative analyses. In the absence of density dependence, using means in Figure 15 with natural mortality, fall abundance should have been reduced by ~ 10% during 1995-2005. This would have an equivalent effect of reducing the summer-fall survival index by 10%. This would have made little difference to fall abundance in the context of the approximately 50-fold variation in summer-fall survival (Figure 17), and would be difficult to detect through correlation.

Although summer-fall survival appears to dominate variability in abundance of delta smelt in fall (Figure 17), this does not imply that control of export effects would be fruitless, as these effects can be considerable during dry years. Management of delta smelt should incorporate any opportunities that arise to improve habitat or food supply and to reduce any negative impacts of predation or toxic contamination. However, current evidence does not provide a clear path toward improving the status of delta smelt using these factors. Manipulating export flow (and, to some extent, inflow) is the only means to influence the abundance of delta smelt that is both feasible and supported by the current body of evidence, even though export effects are relatively small. The results presented here can be used to suggest when, and under what conditions, control of export effects would be most helpful.

AR 018878. Kimmerer (2008) concludes that even though correlative analysis revealed “no effect of export flow on subsequent midwater trawl abundance,” there is reason to be concerned about episodic effects caused by “large proportional losses of delta smelt that occur in some years.”

Id.

As a result, according to Kimmerer (2008), population level effects should be calculated, rather than inferred from correlative analysis.

Id.

After performing such a calculation, Kimmerer (2008) concluded that entrainment reduced “the summer-fall survival index by 10%” during 1995-2005.

Id.

Although this 10% figure was small in the context of the 50-fold variation in summer-fall survival, Kimmerer (2008) nonetheless recommended controlling export effects on smelt because “Manipulating export flow (and to some extent, inflow) is the only means to influ

*879

ence the abundance of delta smelt that is both feasible and supported by the current body of evidence, even though export effects are relatively small.”

Id.

(emphasis added).

Dr. Manly is correct that Kimmerer (2008) does not support the position that entrainment has a “sporadically significant” effect on delta smelt abundance from one year to the next. However, contrary to Dr. Manly’s suggestion, the BiOp does not rely on Kimmerer (2008) for this premise. The BiOp qualifies its reliance on Kimmerer (2008), consistent with the narrow scope of Kimmerer’s findings:

The population-level effects of delta smelt entrainment vary; delta smelt entrainment can best be characterized as a sporadically significant influence on population dynamics. Kimmerer (2008) estimated that annual entrainment of the delta smelt population (adults and their progeny combined) ranged from approximately 10 percent to 60 percent per year from 2002-2006. Major population declines during the early 1980s (Moyle et al. 1992) and during the recent POD years (Sommer et al. 2007) were both associated with hydrodynamic conditions that greatly increased delta smelt entrainment losses as indexed by numbers of fish salvaged. However, currently published analyses of long-term associations between delta smelt salvage and subsequent abundance do not support the hypothesis that entrainment is driving population dynamics year in and year out ( Bennett 2005 ; Manly and Chotkowski 2006; Kimmerer 2008).

BiOp at 210 (emphasis added). It was not unreasonable for FWS to rely on Kimmerer (2008) to conclude that salvage events may be “sporadically significant.” Plaintiffs’ argument that FWS misinterpreted Kimmerer (2008) is unfounded. Kimmerer (2008) explains why, despite the absence of a statistically significant correlation between export pumping and the subsequent year’s smelt population (i.e., between export pumping and the population growth rate), the demonstrated “sporadically significant” loss of smelt within year classes could significantly contribute to the species’ jeopardy. FWS reasonably relied on Kimmerer (2008) for this finding.

Applying Kimmerer’s estimates of entrainment and other data, the BiOp analyzed the effect Project operations have on the frequency of relatively large loss events. For larval and juvenile delta smelt:

Kimmerer (2008) proposed a method for estimating the percentage of the larval-juvenile delta smelt population entrained at Banks and Jones each year. These estimates were based on a combination of larval distribution data from the 20-mm survey, estimates of net efficiency in this survey, estimates of larval mortality rates, estimates of spawn timing, particle tracking simulations from DWR’s DSM-2 particle tracking model, and estimates of Banks and Jones salvage efficiency for larvae of various sizes. Kimmerer estimated larval-juvenile entrainment for 1995-2005. We used Kimmerer’s entrainment estimates to develop multiple regression models to predict the proportion of the larval-juvenile delta smelt population entrained based on a combination of X2 and OMR....

BiOp at 220. The BiOp predicts that “the proposed action will decrease the frequency of years in which estimated entrainment is [less than or equal to] 15 percent. Thus, over a given span of years, the project as proposed will increase larval-juvenile entrainment relative to 1995-2005 levels. This will have an adverse

*880

effect on delta smelt based on their current low population levels.” BiOp at 222.

For adult delta smelt:

The median OMR flows from the CAL-SIM II modeled scenarios were more negative than historic OMR flow for all WY types except critically dry years (Figure E-3; see Table E-5b for all differences). Overall, proposed OMR flows are likely to generate increases in population losses compared to historic years (Figure E-5 and Figure E-6). For example, the frequency of years when population losses are less than 10 percent from most modeled studies (except studies 7.0 and 8.0) is less than 24 percent compared to historic estimates that only exceed 10 percent in approximately half of the years.

The most pronounced differences occur during wet years, where median OMR flows are projected to be approximately 400 to 600 percent (-7100 to -3678 cfs) higher than historical wet years (-1032 cfs). Generally, wet years are marked by low salvage and population losses. However, the proposed operations during wet year are predicted to cause up to a 65 percent increase in smelt salvage and lower probability that population losses will be below 10 percent.

The proposed operation conditions likely to have the greatest impact on delta smelt are those modeled during above normal WYs. The modeled OMR flows for the above normal WYs ranged between -8155 and -6242 cfs, a 33 to 57 percent decrease from the historic median of -5178 cfs. Though the predicted salvage would only be about 15-20 percent higher than historic salvage during these years (Table E-5c), the modeled OMR flows in these years would increase population losses compared to historic years.

In below normal and dry WYs, proposed OMR flows are also modeled to decrease from historic medians. Predicted salvage levels are likely to increase between 2 and 44 percent. More importantly, the modeled median flows from all studies in these WY types range between -5747 and -7438 cfs. Modeled OMR flows at these levels are predicted to increase salvage and increase the population losses from historic levels as well.

During critically dry years, the median OMR flows for studies 7.0, 7.1, 8.0, 9.1, 9.4, and 9.5 are less than -5,000 cfs. These studies have predicted salvage lower than historic salvage and are not likely to generate larger population losses compared to historic years. The models might overestimate salvage during critical dry years when smelt are unlikely to migrate towards the Central Delta due to lack of turbidity or first flush. Thus, the effects of critical dry operations on delta smelt take are probably small and lower than estimated.

In summary, adult entrainment is likely to be higher than it has been in the past under most operating scenarios, resulting in lower potential production of early life history stages in the spring in some years. While the largest predicted effects occur in Wet and Above Normal WYs, there are also likely adverse effects in Below Normal and Dry WYs. Only Critically Dry WYs are generally predicted to have lower entrainment than what has occurred in the recent past.

BiOp at 212-13.

This approach is consistent with Kim-merer (2008). The BiOp does not focus on whether there is a statistically significant correlation between OMR flows and the

*881

population growth rate.

10

Rather, following Kimmerer (2008), the BiOp focuses on predicting the frequency of large salvage events and concluded that Project operations increase their frequency. It was not arbitrary, capricious, or clear error for FWS to base its jeopardy conclusion in part on these predictions of relative increases in entrainment.

See

BiOp at 276.

b.

Population Level Analysis/Life-Cycle Modeling.

Plaintiffs maintain the BiOp’s failure to employ a life-cycle model ignored the best available science. Doc. 551 at 21-22. Using a quantitative

11

life-cycle model

12

is a recognized (the best) method to evaluate the effects of an action upon a fish population’s growth rate. Dr. Richard B. Deriso

13

opined that a population growth rate analysis is the generally accepted method utilized by fisheries biologists to evaluate the impact of a stressor on a fish species’ population. Declaration of Dr. Richard B. Deriso, Doc 401, at I 36;

see also

Declaration of Dr. Ray Hilborn

14

, Doc. 393, at ¶¶ 7-16 (agreeing that life-cycle models are the accepted method in population dynamics to evaluate anthropogenic effects on the probability of growth or decline of a species); Declaration of Ken B. Newman

15

, Doc. 484, at I 8 (agreeing with “utility of life history models for assessing population level effects of SWP/ CVP operations.”). Dr. Hilborn explained that a

quantitative

population dynamics/life cycle model can help distinguish human actions that have a significant impact on population size from those that have little impact on population size, because competition for a resource that is independent of the human activity may cause significant mortality at one stage in the species’ life cycle, meaning that human actions that kill fish at that life stage may have little impact on the population level later in the life history. Hilborn Deck, Doc. 393 at ¶ 15.

Federal Defendants knew of the value of life-cycle modeling. At a March 8, 2007 meeting on the OCAP ESA Re-consultation, attended by FWS employees, the importance of using a life cycle model was emphasized and inquiry made about the progress to date. AR 016016-016017. During the Delta Smelt Action Evaluation Team meeting on August 8, 2008, that Team recognized that population models for delta smelt already had been developed, and that those models were a starting point for quantitative analyses when

*882

combined with appropriate assumptions. AR 011381-011382;

see also

AR 010023, 010027-010029.

There is considerable dispute over whether an appropriate life-cycle model (i.e., one sufficient to perform the types of analyses that would be helpful in the BiOp) existed at the time the BiOp issued. Dr. Newman declares:

Despite the utility of life history models and despite the information that the various surveys provide about different life history stages, an adequately realistic quantitative delta smelt life history model that has been fit using fish survey data does not exist. The BiOp did in many places (e.g., pp. 146, 184, 203) consider the full life history of delta smelt but considerations were via conceptual models in contrast to quantitative models with parameters estimated from data. Part of the difficulty is that there are currently no off-the-shelf computational programs for fitting such a model to data and one must develop customized, computer intensive software. The need to model the spatial and temporal changes in population abundances and to account for the different sources of uncertainty makes model formulation and fitting complex. In particular, uncertainty in survey data, due to random sampling error and bias, complicates model fitting. Capture probabilities differ between surveys, the probabilities are largely unknown (despite efforts made to estimate them, for example, for FMWT data, see Newman 2008 (Administrative Record “AR” at 19782-19799)), and capture and fish presence probabilities are thus confounded. Furthermore, given the patchiness and heterogeneity of the spatial and temporal distribution of delta smelt and the relatively low capture probabilities (whatever they might be), the sampling errors associated with survey data can be quite large (Newman 2008 (AR at 19782-19799)). Failure to account for sampling errors may result in biased parameter estimates (including wrongly concluding density dependence; Shenk et al. 1998).

The difficulties are not insurmountable, but concentrated research efforts are required.

I know of three such efforts currently underway and at varying stages of development: (1) an individual-based model with a spatial component by Drs. Wim Kimmerer, San Francisco State University, William Bennett, University of California at Davis, Stephen Monismith, Stanford University, and Kenneth Rose, Louisiana State University; (2) a population-level life history model using information from multiple surveys by Dr. Mark Maunder, Inter American Tropical Tuna Commission; (3) similar to Maunder, a life history model with a spatial component based on multiple surveys’ data has been conceptually sketched by me and others in the NCEAS POD working group. Given sufficient time and appropriate technical resources, including personnel, to focus on model formulation and fitting, these models might be available within a year.

Newman Dec!., Doc. 484 at ¶ 5.

All of the experts agreed with Dr. Newman that, at the time the BiOp was issued, there was no “off-the-shelf’ life-cycle model to apply to delta smelt. Considerable dispute exists over how long it should have taken FWS to develop a competent model. It is undisputed that basic life-cycle models such as the Ricker model can be applied to fisheries data sets in relatively short order. Deriso Decl., Doc. 605, at ¶ 52. Dr. Deriso opined that FWS had all the data necessary to perform a life-cycle analysis. Deriso Deck, Doc. 401, at ¶70. Dr. Hilborn stated that a relatively complex life-cycle model that “follow[s] the

*883

size structure of delta smelt through then-life history and fit this into the observed size structure” would “require no more than a few months time to construct, evaluate and use in a biological opinion.” Hilborn Decl., Doc. 600 at I 14. Dr. Punt, a 706 Expert with expertise in fish population dynamics and biostatistics,

see

Doc. 394 at 2, stated “[i]t is surprising that a population dynamics model was not developed for delta smelt for the BiOp.... The model developed by Bennett could have been extended to more fully account for the biology of delta smelt and fitted to data to assess the population-level effects of impact of the project.” Doc. 633-1 at 3.

Federal Defendants’ expert, Mr. Feyrer disagrees:

Developing a quantitative population model is a challenging and complex exercise that could not have been completed by USFWS within the timeframe required to issue the 2008 BiOp. The work requires a substantial investment of resources and individuals with very specialized skills. The process to develop, test, peer-review, and apply such models often takes years. For instance ... the development of models for Columbia River salmon ... took no less than three years to complete.

Because of the recognized urgent need for such tools, there are on-going efforts to develop quantitative population models for delta smelt. For instance, Bennett (2005) presented preliminary results from a stage-structured model he is developing to examine tradeoffs among sources of mortality acting on different cohorts and life stages.

See

AR at 17004-74. The development of this model is part of a broader comprehensive effort by a team of researchers including Dr. Kenneth Rose of Louisiana State University, Dr. Wim Kimmerer of San Francisco State University, Dr. William Bennett of the University of California at Davis, and Dr. Stephen Monismith of Stanford University, who are in the early stages of developing, testing, and applying particle-tracking models, an individual-based model, and a matrix projection model. The development of these particular models is very promising but has also been faced with many challenges. Perhaps the most critical challenge has been a freeze on project funding by the State of California; it is uncertain if the funding will be reinstated. Another example is the work I have been personally involved with at NCEAS. The NCEAS team has used Bayesian changepoint techniques and multivariate autoregressive modeling to identify factors contributing to the decline of delta smelt and other species. The results of this work will be published in two papers in an upcoming issue of the journal Ecological Applications. I am aware of at least two other independent efforts of modeling the effects of various stressors on delta smelt that are also under development. Unfortunately, none of the work I mention above was available when the 2008 BiOp was being prepared. To my knowledge, no comprehensive quantitative population dynamics model for the delta smelt has been developed, subjected to peer-review, and published.

... [Quantitative population models are grounded in what is known about the biology of a species, and processes that may plausibly affect its abundance.... Although there is a substantial amount of data available on delta smelt, a key problem is that much of the sample data has increasingly contained zero values. These zeros are a reflection of declining population abundance. Such low numbers make it more difficult to acquire more recent information about the factors that drive delta smelt population

*884

dynamics, such as survival probabilities by life history stage, movement patterns and spatial distribution, and fecundity or reproductive success. It is thus becoming increasingly difficult to not only simply estimate such factors, but also increasingly difficult to model how these factors are affected by environmental and anthropogenic processes such as those considered in the 2008 BiOp. The estimation of delta smelt population size exemplifies this problem. Newman (2008), see AR at 19782-99, recently published a sample design-based procedure for estimating the population abundance of pre-adult and adult delta smelt. However, the resulting estimates of population size were quite imprecise. This was caused, in part, by limitations of the available data to estimating capture probabilities and gear efficiency.

... I agree ... that population dynamics models have been used to evaluate consequences of various stressors on a wide range of species and human impacts. I also agree that there is sufficient data to develop such a model for delta smelt, as demonstrated by the examples I provided above. However, although some are in development, the fact remains that no such model has been fully developed, peer-reviewed and made available for application. Thus, in the absence of such models, I disagree that that the techniques used by USFWS were inconsistent with generally-accepted scientific standards and practices. To the contrary, in the absence of such a model, and because one could not be developed during the time allowed for this consultation, the techniques used by USFWS do reflect generally-aecepted scientific standards and practices.

Decl. of Frederick V. Feyrer

16

, Doc. 541, at ¶¶ 30-33. Plaintiffs do not suggest any party that participated in the preparation of the OCAP Biological Assessment (“OCAP BA” or “BA”) or commented on the public review drafts of the BiOp during the consultation submitted to FWS a quantitative life cycle model or the results of such an analysis using a life cycle model for delta smelt.

The ESA does not require FWS’s to generate new studies. In

Southwest Center for Biological Diversity v. Babbitt,

215 F.3d 58 (D.C.Cir.2000), the district court found “inconclusive” the available evidence regarding FWS’s decision not to list the Queen Charlotte goshawk, and held that the agency was obligated to find better data on the species’ abundance. The D.C. Circuit reversed, emphasizing that, although “the district court’s view has a superficial appeal ... this superficial appeal cannot circumvent the statute’s clear wording: The secretary must make his decision as to whether to list a species as threatened or endangered’ solely on the basis of the best scientific and commercial data available to him....’ 16 U.S.C. § 1533 (b)(1)(A).”

Id.

at 61 (emphasis added);

see also American Wildlands v. Kempthorne,

530 F.3d 991, 998 (D.C.Cir. 2008) (the “best available data” standard “requires not only that the data be attainable, but that researchers in fact have conducted the tests”).

Plaintiffs advocate a narrow reading of both

Southwest Center

and

American Wildlands ,

arguing these cases only mean that the agency is not required to gather new data in the field regarding a species if such information is not already available.

*885

Doc. 697 at 22. Plaintiffs object that “[njeither of these cases supports Defendants’ position that FWS could disregard the smelt abundance data that were already in its possession and fail to undertake the necessary statistical analyses to satisfy its statutory mandate to determine’ whether the action ... is likely to jeopardize the continued existence of the species.’ 50 C.F.R. § 402.14 (g)(4).”

Id.

Plaintiffs cite no authority suggesting that the non-existence of an analytical model should be treated any differently from the non-existence of raw field data. FWS did not have an off-the-shelf form of “statistical analysis” it could apply to determine the effects of Project Operations on the delta smelt population. Although life-cycle modeling is standard practice in the field of fisheries biology, and a life-cycle model is being (and should have been) developed for delta smelt, it is undisputed that an appropriate life cycle model had not been developed at the time the BiOp issued. FWS must apply the best “available” science; not the best science possible. FWS’s failure to apply a life cycle model did not per se violate the ESA or the APA.

It is undisputed that application of a quantitative life cycle model is the preferred scientific methodology. Based on the preponderating expert testimony, FWS had the time and ability to prepare the necessary life-cycle model. FWS made a conscious choice not to use expertise available within the agency to develop one. A court lacks authority to require completion of a life-cycle model. In light of uncontradicted expert testimony that life-cycle modeling is necessary and feasible, FWS’s failure to do so is inexplicable,

c.

FWS’ Use of Raw Salvage Numbers.

Plaintiffs argue that FWS’s use of raw salvage numbers in its quantitative justification for the flow prescriptions in Actions 1 and 2 constitutes a failure to apply the best available science. Action 1, designed to protect upmigrating delta smelt, is triggered during low and high entrainment risk periods based on physical and biological monitoring. Action 1 requires OMR flows to be no more negative than -2,000 cubic feet per second (“cfs”) on a 14-day average and no more negative than -2,500 cfs for a 5-day running average. BiOp at 280-81, 329-30. Action 2, designed to protect adult delta smelt that have migrated upstream and are residing in the Delta prior to spawning, is triggered immediately after Action 1 ends or if recommended by the Smelt Working Group (“SWG”). Flows under Action 2 can be set within a range from -5,000 to -1,250 cfs, depending on a complex set of biological and environmental parameters.

Id. at

281-82, 352-56.

The BiOp provides a quantitative justification for these specific flow prescriptions in Attachment B, entitled “Supplemental Information related to the Reasonable and Prudent Alternative.” The following subsection entitled, “Justification for Flow Prescriptions in Action 1,” is critical to the present challenge and is reproduced here in its entirety:

Justification for Flow Prescriptions in Action 1

Understanding the relationship between OMR flows and delta smelt salvage allows a determination of what flows will result in salvage. The OMR-Salvage analysis herein was initiated using the relationship between December to March OMR flow and salvage provided by P. Smith and provided as Figure B-13, below. Visual review of the relationship expressed in Figure B-13 indicates what appears to be a “break” in the dataset at approximately -5,000 OMR; however, the curvilinear fit to the data suggest that the break is not real and

*886

that the slope of the curve had already begun to increase by the time that OMR flows reached -5,000 cfs.

[[Image here]]

'Figure B 13. OMR Salvage relationship for adult delta smelt, (source,

P.

Smith). Data from tills figure were the raw data used in the piecewise polynomial regression analysis.

Further, a nonlinear regression was performed on the dataset, and the resulting pseudo-R2 value was 0.44—suggesting that although the curvilinear fit is a reasonable description of the data, other functional relationships also may be appropriate for describing the data. Fitting a different function to the data could also determine the location where salvage increased, i.e. identify the “break point” in the relationship between salvage and OMR flows. Consequently, an analysis was performed to determine if the apparent break at - 5,000 cfs OMR was real. A piecewise polynomial regression, sometimes referred to as a multiphase model, was used to establish the change (break) point in the dataset.

A piecewise polynomial regression analysis with a linear-linear fit was performed using data from 1985 to 2006. The linear-linear fit was selected because it was the analysis that required the fewest parameters to be estimated relative to the amount of variation in the salvage data. Piecewise polynomial regressions were performed using Number Cruncher Statistical Systems ((c) Hintz, J., NCSS and PASS, Number Cruncher Statistical Systems, Kaysville UT).

The piecewise polynomial regression analysis resulted in a change point of - 1162, i.e. at -1162 cfs OMR, the slope changed from 0 to positive (Figure B-14). These results indicate that there is a relatively constant amount of salvage at all flows more positive than -1162 cfs

*887

but' that at flows more negative than - 1162, salvage increases. The pseudo-R2 value was 0.42, a value similar to that obtained by P. Smith in the original analysis.

To verify that there was no natural break at any other point, the analysis was performed using a linear-linear-linear fit (fitting two change points). The linear-linear-linear fit resulted in two change points, -1,500 cfs OMR and - 2,930 cfs OMR. The -1,500 cfs value is again the location in the dataset at which the slope changes from 0 to positive. The pseudo-R2 value is 0.42 indicating that this relationship is not a better description of the data. Because of the additional parameters estimated for the model, it was determined that the linear-linear-linear fit was not the best function to fit the data, and it was rejected. No formal AIC analysis was performed because of the obvious outcome.

A major assumption of this analysis is that as the population of Delta smelt declined, the number of fish at risk of entrainment remained constant. If the number of fish in the vicinity of the pumps declined, fewer fish would be entrained and more negative OMR flows would result in lower salvage. This situation would result in an overestimate, i.e. the change point would be more positive. In fact, if the residuals are examined for the relationship in Figure B-13 above, the salvage for the POD years 2002, 2004, 2005, and 2006 are all below the line. 2003 is above the line although the line is not extended to the points at the top of the figure, and these data points occur when the curve becomes almost vertical. The negative residuals could be a result of a smaller population size available for entrainment and salvage. This could be verified by normalizing the salvage data by the estimated population size based on the FMWT data.

[[Image here]]

Figure B-14. Piecewise polynomial regression of OMR flows and salvage. The change point is the location at which the two regression lines meet; -1,162 cfs OMR,

The original values of OMR and salvage could have been measured with error due to a number of causes, consequently the values used in the original piecewise

*888

polynomial analysis could be slightly different than the “true” values of salvage and OMR flow. Consequently, a second analysis was undertaken to examine the effect of adding stochastic variation to the OMR and salvage values in the piecewise polynomial regression analysis. The correlation between OMR and salvage in the original dataset was -0.61 indicating that the more negative the OMR, the greater the salvage. Consequently, it was necessary to maintain the original covariance structure of the data when adding the error terms and performing the regressions. The original covariance structure of the OMR—salvage data was maintained by adding a random error term to both parameters. The random error term was added to OMR and a correlated error term was added to salvage. The expected value of the correlated errors was -0.61.

The error terms were selected from a normal distribution with a mean of 1.0 and a standard deviation of 0.25 which provided reasonable variability in the original data. Operationally this process generated a normal distribution of OMR and salvage values in which the mean of the distributions were the original data points. Additional analyses were performed with standard deviations of 0.075, 0.025, and 0.125. Smaller standard deviations in the error term resulted in estimates of the change point nearer to the original estimate of -1,162 cfs. This is to be expected as the narrower the distribution of error terms, the more likely the randomly selected values would be close to the mean of the distribution. The process was repeated one hundred times, each time a new dataset was generated and a new piecewise polynomial regression was performed. The software package @Risk ((c) Palisade Decision Tools) was used to perform the Monte Carlo simulations. Latin hypercube sampling was used to insure that the distributions of OMR and salvage values were sampled from across their full distributions. The parameter of interest in the simulations was the change point, the value of the OMR flow at which the amount of salvage began to increase. Incorporating uncertainty into the analysis moved the change point to -1,800 cfs OMR, indicating that at flows above -1683, the baseline level of salvage occurred but with flows more negative than -1683, salvage increased.

BiOp 347-51 (emphasis added).

The analyses contained in Figures B-13 and B-14 serve,

inter alia,

as justification for Action 1: setting “break points” above and below which entrainment rates noticeably change. These break points are the foundation for the tiered flow restrictions in RPA Action 1. Cay Collette Goude

17

stated in her expert declaration that the analysis conducted by Dr. Michael Johnson, set forth in Figure B-13, found inflection points where entrainment started to increase with more negative OMR flows, and that the inflection point “was -1,800 cfs OMR when uncertainty was factored into the analysis.” Doc. 470, at ¶22. The BiOp does not explain in the “Justification for Flow Prescriptions in Action 1” or elsewhere how or why this -1,800 cfs figure relates to the -2,000 cfs upper limit imposed by Action l.

18

*889

Action 2 calls for flows to be set within a range from -5,000 to -1,250 cfs, depending on a complex set of biological and environmental parameters. BiOp

at

281-82, 352-56. Although Appendix B describes and justifies Action 2 separately from Action 1, there is no independent section justifying the flow prescriptions imposed by Action 2. Instead, there is a subsection entitled “Justification for Guidelines in Setting Prescriptions of Action 2” which fixes biological and environmental parameters the SWG is to use in setting flows within the - 5,000 cfs to -1,250 cfs range.

See

BiOp at 355. There is no independent quantitative or qualitative justification for the upper and lower limits of that range. In fact, the “Justification for Guidelines in Setting Prescriptions of Action 2” section contains the following statement:

Flow requirements defined within Action 2 follow the same protectiveness criterion established during Action 1, as adjusted to reflect real-time conditions and predicted entrainment risk relative to the anticipated distribution and abundance of year-class delta smelt; and reflecting their behavioral propensity to hold in their chosen spawning habitat. These are allowed to vary based upon assessment of available data as described in the adaptive process described in the Introductions to Actions section above.

BiOp at 356.

Plaintiffs complain that the “Justification for Flow Prescriptions in Action 1” section does not represent the best available science because it is based upon analyses of gross (or “raw”) salvage (i.e. the absolute number of fish salvaged over a given time period). The use of raw salvage data, as opposed to salvage data scaled to population size, is problematic because raw salvage figures do not account for the size (or relative size) of the smelt population. Deriso Decl., Doc. 401, at ¶ 28. The BiOp admits as much, and concedes that the analysis assumes that “as the population of Delta smelt declined, the number of fish at risk of entrainment remained constant.” BiOp at 349. Considering raw salvage numbers alone provides no means of distinguishing an event in which 10,000 ñsh are salvaged out of a population of 20,000 from an event in which 10,000 fish are salvaged from a population of 20 million. Deriso Decl., Doc. 401, ¶ 28.

There is widespread agreement among the scientific experts that the use of normalized salvage data rather than gross salvage data is the standard accepted scientific methodology among professionals in the fields of fisheries biology/management. Doc. 633-1 at 7, 10 (the 706 experts concluded that, although it is not inherently unreasonable to consider the analysis in Figure B-13, it would be unreasonable to rely on that analysis as the only basis for imposing flow restrictions); Deriso Deck, Doc. 401 at ¶¶ 51-56 (FWS’s reliance on Figure B-13 to conclude that as negative OMR flows increase, more adults are salvaged is “scientifically flawed because raw salvage numbers do not have a directly proportional effect on population and do not take into account the overall size of the population....”); Newman Decl., Doc. 484 at ¶ 11 (concurring with Dr. Deriso’s “general notion of scaling salvage by some measure of population size.”).

*890

FWS was aware that raw salvage data posed this obvious problem. The BiOp itself recognized the necessity of normalizing raw salvage data:

To provide context to determine the magnitude of effect of pre-spawning adult direct mortality through entrainment within any given season (as measured by salvage), it is necessary to consider two important factors ¶ The second factor to consider when relating salvage to population-level significance is that the total number salvaged at the facilities does not necessarily indicate a negative impact on the overall delta smelt population.

BiOp at 388. The August 26, 2008, draft meeting notes of FWS’s Delta Smelt Action Evaluation Team state:

When analyzing the importance of entrainment to the species population structure or decline, the relevant fact to consider is the percentage of the population being removed via entrainment. Salvage data, by itself, may not be sufficient to help one understand the percentage of the population being removed via entrainment.

AR 010023. The Independent Peer Review of FWS’s draft Effects Analysis for the BiOp also recommended to FWS that it “normalize[ ]” salvage to population size:

The panel suggests that the use of predicted salvage of adult smelt should be normalized for population size. Total number salvaged is influenced by a variety of factors, particularly the number of fish in the population.... Expressing salvage as a normalized index may help remove some of the confounding of the temporal trends during the baseline.

AR 008818. FWS used normalized salvage data in other parts of the BiOp, including the calculation of the Incidental Take Limit, evidencing its ability to do so.

See

Deriso Decl., Doc. 401, at ¶ 55 (citing BiOp at 386).

FWS nowhere explains its decision in the BiOp to use gross salvage numbers in Figures B-13 and B-14, and does not explain why it selectively used normalized salvage data in some parts of the BiOp but not in others.

See

Doc. 633-1 at 10 (Dr. Thomas Quinn, a 706 Expert with expertise in fisheries biology, estuarine ecology, and fish migration and movement,

see

Doc. 394 at 2, stated: “it is not clear why such an adjustment [of salvage to population size] was not made for the data examined in this report.”). This was arbitrary, capricious, and represents a failure to utilize the best available science in light of universal recognition that salvage data must be normalized. This significant error must be corrected on remand.

(1)

Federal Defendants’ Argument that the Flow Prescriptions in Actions 1 and 2 are Otherwise Justified.

Federal Defendants argue that the specific flow prescriptions in Actions 1 and 2 are supported by more than just Figures B-13 and B-14. By portraying a negative as a positive, Federal Defendants point out that nothing in the BiOp suggests Figures B-13 and B-14 are in fact being used to draw conclusions about what is happening to the delta smelt population as a whole. Doc. 660 at 32. The BiOp concedes that “when relating salvage data to population-level significance [ ] the total number salvaged at the facilities does not necessarily indicate a negative impact upon the overall delta smelt population.” BiOp at 338. Instead, Federal Defendants suggest that the raw salvage numbers are used in “tandem” with other population-based analyses. Other sections of the BiOp demonstrate that salvage by the Project pumping facilities can have a “sporadically significant” effect on the delta smelt population.

*891

However, Federal Defendants concede that neither the research supporting the “sporadically significant” finding nor any related discussion in the BiOp generate the kind of “operational metric ... needed so that Project pumping can be managed to prevent the entrainment numbers that these other population analyses deem necessary for avoiding population level effects.” Doc. 660 at 32-33. Federal Defendants argue that the ráw salvage analyses contained in Figures B-13 and B-14 are used solely to generate these “operational metrics”:

That is where raw salvage comes in—it works in tandem with these other population-based analyses, which Plaintiffs disregard. Specifically, Figures B-13 and B-14 are included to illustrate that the Projects quickly lose the ability to manage entrainment and salvage risk once OMR flows become more negative than -5000 cfs. This is the level at which it is believed that entrainment losses or the take level can be effectively managed.

See

BiOp at 366 (explaining that the function of the OMR flow targets is to manage entrainment risk).

Id.

at 33. This argument does absolutely nothing to overcome the fact that the use of raw salvage in the analyses depicted in figures B-13 and B-14 is scientifically unacceptable. Those figures cannot accurately depict when the Projects “lose the ability to manage entrainment and salvage risk,” because they do not scale salvage to population size. These figures do not take into account the possibility that one data point used to generate the curves depicted may have been collected in a year when the delta smelt population was 1,000,000, making it more likely that larger numbers of smelt would be present near the pumps to be salvaged, while another data point might have been collected during a year in which the population was 10,000, making it inherently less likely that large numbers of smelt would be found in salvage. The present record suggests that such metrics are meaningless as management tools. They cannot be used to set specific flow prescriptions. FWS was offered the opportunity to, but has not justified its approach.

At the same time, Federal Defendants contend that at least some of the “break points” reflected in the specific flow prescriptions of Components 1 and 2 are based on information unrelated to Figures B-13 and B-14. For example, in the justification for Action 3, which is designed to protect larval & juvenile smelt, the BiOp relies upon Particle Tracking Model (“PTM”)results to explore the likelihood of entrainment of particles in the south Delta (used to represent that portion of the smelt population located in the south Delta) that would likely be entrained at various levels of negative OMR flow. This is referenced as “entrainment risk”:

The most efficient protective measure for protecting the resilience and not precluding the recovery of the delta smelt population specific to the larval/juvenile lifestage is to prevent entrainment of fish in as large a portion of the Central Delta as is practical. Results of PTM modeling focusing on protections at station 815 (Prisoner’s Point) indicates that precluding entrainment of larval/juvenile delta smelt at this station would also protect fish at station 812 (Fisherman’s Cut) and other stations north and west (downstream) of station 815. While the target entrainment at station 815 would ideally also be zero, there appears to be little additional entrainment protection (less than 5 percent) at OMR flows at - 750 cfs (the strictest level addressed by Interim Remedies). However, entrainment risk grows exponentially at OMR flows increasingly more negative than - 2000 cfs.

*892

Figure B-16 displays injection points for modeled particle tracking runs that were conducted in February 2008 with injection points at Stations 711, 809, 812, 815, 902, 915. This figure plots projected relationships for OMR flows by injection point, including entrainment probabilities for station 815 (over 30 days).

The results from these runs indicate an approximate << 5 percent entrainment risk at OMR flow not more negative than -2000 cfs. At a requirement of - 3,500 cfs OMR flow, entrainment risk at station 815 is roughly 20 percent over each 30 day interval. Assuming cumulative entrainment is additive, over a roughly four month (~120 days) interval in which Action 3 would be under effect, consistently operating at -3,500 OMR would yield a net entrainment probability placing at risk approximately 80 percent of the larval/juvenile subpopulation utilizing the South Delta at and below Station 815. If immigration of larval smelt from the Central or North Delta into the zone of entrainment during spring were to occur, the population-level risk would be even greater. Such entrainment levels are potentially a significant adverse risk to delta smelt population.

BiOp at 366-68.

Although it seems logical that the PTM results and the “entrainment risk” PTM attempts to estimate have some applicability to the protection of adult smelt, the BiOp does not rely upon these results to justify Actions 1 or 2.

NWF v. NMFS II,

524 F.3d at 932, n. 10 (a court “may not consider [a] post hoc justification, or infer ‘an analysis that is not shown in the record.’ ”) (quoting

Gifford Pinchot Task Force,

378 F.3d at 1074, and citing

PCFFA v. U.S. Bureau of Reclamation,

426 F.3d 1082, 1091 (9th Cir.2005) (“[W]e cannot infer an agency’s reasoning from mere silence,” and “an agency’s action must be upheld, if at all, on the basis articulated by the agency.”)).

Federal Defendants also point out that Action 1 is based on “the historical observation that the first ‘winter flush’ moves delta smelt into portions of the delta where they are particularly vulnerable to entrainment, for biological and hydrological reasons that are well documented.” Doc. 660 at 23 (citing BiOp at 333-36). Federal Defendants argue:

As the multiple sources of information relied upon by the BiOp on this point demonstrate, pumping reductions during these critical vulnerability periods will demonstrably reduce entrainment and entrainment risk.

See id.

According to the BiOp, the piece-wise regression set forth in Figure B-14 of the BiOp was used to provide some indication of what level of exports would reduce entrainment during these first flush events, and not, as Plaintiffs assert, to analyze the impacts of salvage relative to the population.

See

BiOp at 350.

Doc. 660 at 23. The BiOp arguably supports the assertion that a “winter flush” can move smelt into areas of the delta where they are particularly vulnerable.

See

BiOp at 331. However, nothing in the discussion of the timing, characteristics, or indicators of the winter flush explains why -5,000 cfs was set as the ceiling on negative OMR flows, rather than some other figure. That justification appears to come exclusively from Figures B-13 and B-14, which rely upon the flawed analyses of raw salvage.

Finally, Federal Defendants attempt to justify the use of raw salvage numbers in calculating the -5,000 cfs ceiling by a convoluted argument that Kimmerer’s work proves raw salvage trends generally follow population trends. Kimmerer’s work did

*893

evaluate the population-level effects of project operations. The BiOp explains:

This effects analysis evaluates the proposed action operations by exploring long-term trends in Delta outflow, or X2, and OMR flows during March-June and comparing these to hydrodynamic conditions expected based on CALSIM II modeling presented in the biological assessment. The analysis uses the larval-juvenile entrainment estimates provided by Kimmerer (2008) and flow and export projections from the biological assessment to estimate the annual percentages of the larval/juvenile delta smelt population expected to be entrained ....

Kimmerer (2008) proposed a method for estimating the percentage of the larval-juvenile delta smelt population entrained at Banks and Jones each year. These estimates were based on a combination of larval distribution data from the 20-mm survey, estimates of net efficiency in this survey, estimates of larval mortality rates, estimates of spawn timing, particle tracking simulations from DWR’s DSM-2 particle tracking model, and estimates of Banks and Jones salvage efficiency for larvae of various sizes. Kimmerer estimated larval-juvenile entrainment for 1995-2005. We used Kimmerer’s entrainment estimates to develop multiple regression models to predict the proportion of the larval/juvenile delta smelt population entrained based on a combination of X2 and OMR.

BiOp at 219-220 (emphasis added). The BiOp used a similar approach for adult delta smelt:

Kimmerer (2008) calculated that entrainment losses of adult delta smelt in the winter removed 1 to 50 percent of the estimated population and were proportional to OMR flow, though the high entrainment case might overstate actual entrainment. Given there are demonstrated relationships between smelt entrainment and salvage with OMR flows (Kimmerer 2008; Grimaldo et al. accepted manuscript), this effects analysis evaluates the proposed action operations by comparing the long-term trends in OMR flows to OMR flows in the CALSIM II modeling presented in the biological assessment. For both approaches, predictions of salvage and total entrainment losses were made using OMR flow since it was the best explanatory variable of each. The effects of proposed operations were determined by comparing actual salvage and entrainment losses with predictions of these parameters under modeled OMR flows.

BiOp at 211 (emphasis added). Kimmerer did calculate proportional population-level losses for both adults and juveniles. See id;

see also

BiOp at 212, 250-252, 262 (presenting model simulation results in Figures E4-E6 and E16 which estimate proportional population losses based on entrainment). It is undisputed, however, that Kimmerer did not generate any operational metrics or attempt to calculate the point above or below which OMR flows would have particular effects on the smelt population. As a result, there was no basis to rely on Kimmerer’s work alone to justify the specific OMR flows imposed by Actions 1 and 2. Federal Defendants point to a section of the BiOp’s Effect’s Analysis that concludes that because “over a given span of years, the project as proposed will increase larval/juvenile entrainment relative to 1995-2005 levels,” “[tjhis will have an adverse effect on delta smelt based on their current low population levels.” BiOp at 222. However, this conclusion references Figure E-18, which attempts to estimate the likelihood of having an event that would entrain a significant proportion of the smelt population, thereby evaluating the effect of particular circum

*894

stances on the smelt population.

See

BiOp at 264. This language provides no support for Federal Defendants’ assertion that the BiOp connects population level effects to raw salvage figures.

Federal Defendants assert “Kimmerer (2008), like the BiOp, concluded that once raw entrainment numbers approach a certain level, population-level effects will occur.” Doc. 660 at 25 (citing BiOp at 159, 164-65, 210; AR at 18854-18880). Federal Defendants describe this as the “Kim-merer Approach,” and argue:

The Kimmerer (2008) study shows that salvage trends generally follow population loss trends.

See

BiOp at 206-207;

see also

AR at 18854-18880. Salvage data is then used to ascertain the pumping level at which entrainment risk can no longer be managed to a level that prevents harm to the population as a whole. See BiOp at 338. Using the Kimmerer approach, by managing salvage, the BiOp manages population-level losses.

Doc. 660 at 25. This description is not supported by the record. The BiOp does not rely upon Kimmerer (2008) or any other source to conclude that salvage trends generally follow population loss trends. This is FWS’s invention to support its arbitrary flow limit.

FWS nowhere explains in the BiOp or the AR how the sporadically significant population-level effects identified in Kim-merer (2008) factored into the quantitative analysis that led to the -5,000 cfs OMR flow limit imposed in RPA Action 2. Nowhere does the BiOp or the record explain how the analysis in Fig. B-13 “works in tandem” with the purported numeric results of Kimmerer (2008), and nowhere does the BiOp or the record state that Fig. B-13 was intended to create an “operational metric” to manage pumping to avoid “certain raw entrainment numbers.” This is an abdication of the duty to satisfy the basic APA requirement that the agency “articulate[ ] a rational connection between the facts found and the choice made.”

Ariz. Cattle Growers’ Ass’n,

273 F.3d at 1236 .

Federal Defendants argue that, even if FWS had used a scaled salvage index to calculate the OMR flow ceiling, the results would not have been appreciably different. For the purposes of demonstrating the difference between the analysis presented in the BiOp and a population-normalized analysis, Dr. Deriso analyzed the relationship between normalized salvage and OMR flows. He initially concluded that there is “no statistically significant relationship between OMR flows and adult salvage for flows less negative than -6,100 [cfs] at the very least.” Deriso Deck, Doc. 401 at ¶¶ 62-65.

19

Federal Defendants’ expert criticized Dr. Deriso’s alternative analysis in a number of ways, including that Dr. Deriso failed to correct for potentially large sampling errors. Newman Deck, Doc. 484, at ¶ 12. Dr. Newman ran his own analysis, applying a different standard statistical methodology to the same data used by Dr. Deriso, and got different results regarding the “inflection point” where OMR flows had an increasing impact on the population-normalized salvage rate.

Id. &

Ex. C (identifying inflection point at -4,000 cfs, which is within the OMR flow target

*895

ranges established in the BiOp). Ultimately, however, Dr. Newman agreed that an analysis utilizing raw salvage numbers (i.e., not adjusted for relative population size) is scientifically inappropriate.

Id.

at ¶ 11. That other researchers were able to produce generally consistent inflection points through the use of more appropriate statistical methodologies does not excuse FWS’s failure to do so. The difference between a -6,100 cfs ceiling and a -4,000 cfs ceiling is very substantial in the amount of lost annual water supply, with resulting adverse effects on human welfare and the human environment. FWS was required to perform an accurate scientific analysis and justify its ultimate decision regarding the imposition of a water flow ceiling.

20

(2)

Use of Raw Salvage Analyses in Justification for Action 3.

Action 3, which is designed to “[m]inimize the number of larval delta smelt entrained at the facilities by managing the hydrodynamics in the Central Delta ...,” limits net daily OMR flow to no more negative than -1,250 to -5,000 cfs, based on a 14-day running average with a simultaneous 5-day running average within 25 percent of the applicable requirement for OMR. BiOp at 357. Action 3 establishes guidelines the SWG is to use when recommending where to set the OMR flow level within this range.

Id.

The BiOp anticipates that during most conditions, OMR flows will range between -2,000 and -3,500 cfs.

Id.

at n. 10. During certain years of higher or lower predicted “entrainment risk,” flows as low as -1,250 or as high as - 5,000 may be recommended.

Id.

Plaintiffs do not challenge the basis for the low end of the range (-1,250 cfs) or the criteria used to formulate recommendations within the middle of the range. Plaintiffs do argue that the upper end of the range (-5,000 cfs) is based solely on FWS’s raw salvage analysis and should be invalidated.

The BiOp explains in the section of Attachment B addressing Action 3 that “[t]wo scenarios span the range of circumstances likely to exist during Action 3”:

First,

the

low-entrainment risk scenario. There may be a low risk of larval/juvenile entrainment because there has been no evidence of delta smelt in the South and Central Delta or larval delta smelt are not yet susceptible to entrainment.

In this scenario, negative OMR flow rates as high as -5,000 cfs may occur as long as entrainment risk factors permit.

The second scenario, the high-entrainment risk scenario, is one in which either (a) there is evidence of delta smelt in the South and Central Delta from the SKT and/or 20mm survey, or (b) there is evidence of ongoing entrainment, regardless of other risk factors. In this case, OMR should be set to reduce entrainment and/or the risk of entrainment as the totality of circumstances warrant. Usually, if the available distributional information suggests that most delta smelt are in the North or North/Central Delta, then OMR flow can be chosen to

*896

minimize Central Delta entrainment. However, if the distributional information suggests there are delta smelt in the Central or South Delta, then OMR flows will have to be set lower to reduce entrainment of these fish. If delta smelt abundance is low, distribution cannot be reliably inferred. Therefore, the adaptive process is extremely important. The SWG may recommend any specific OMR flow within the specified range above.

BiOp at 358 (underlined emphasis in original; emphasis in italics added). The Action 3 discussion does not provide an independent justification for the choice of - 5,000 cfs as the upper limit for OMR flows under the low entrainment risk scenario. Federal Defendants suggest that the upper limit is justified in the Delta Smelt OCAP Team’s notes, which indicate that “[a]t -5,000 OMR, the model shows 40% entrainment at station 815.” AR 009459. This is a reference to the PTM model results. There are two major problems with Federal Defendants’ reliance on this statement. First, it is contained within a section of the Delta Smelt OCAP Team notes entitled “Actions 1 and 2.” AR 009457-60. Even if this statement was made in reference to Action 3, it does not justify using -5,000 cfs as the upper limit. The PTM study assumed an upper limit of -5,000 cfs and never considered any flow ranges above that. Nor is it made clear why 40% particle entrainment is a rational threshold of significance, as opposed to some lower or higher threshold. In sum, the PTM study does not justify the imposition of -5,000 cfs as an upper limit in Actions 1, 2, or 3.

The “Action # 3” section of the Team’s notes does contain an explanatory statement regarding the source of the -5,000 cfs upper boundary for Action 3: “The -5,000 OMR cap was established by Wanger.” AR 009463;

see also

AR 009462 (“[t]he group discussed the merits of using the - 5,000 OMR per Wanger Order”). It is unclear how FWS can rely directly on a provisional court order, entered as a remedial stopgap measure pending comprehensive scientific analysis, to establish the scientific basis for an RPA. The subject Order was the result of an Interim Remedies proceeding in the challenge to the previous Delta Smelt BiOp. After an evidentiary hearing, it was determined from the then available data that “the number of Delta smelt entrained at the CVP and SWP export facilities begins to rise significantly when negative flows on the OMR exceed approximately -5,000 cfs. [Tr. 641:14-642:5; 725:16-17; DWR Ex. D ¶ 4; DWR Ex. G ¶ 34; SWC Ex. N].”

NRDC v. Kempthorne,

1:05-cv-1207, Doc. 561, Int. Rem. Findings, at ¶ 38. The finding was based on two studies of the relationship between OMR flows and smelt salvage: (1) a nonlinear model presented by Sheila Greene of DWR; and (2) the linear model created by Peter Smith, which became the basis for Figure B-13. Both of these analyses utilized raw salvage data. AR 009251 (Green’s analysis);

see also

1:05-cv-1207, Doc. 399, Decl. of Jerry Johns, Ex. B and C; 1:05-cv-1207, Doc. 419, Decl. of Christina Swanson, at 12, Fig. 8. That raw salvage studies were previously relied upon by the Court, when no others were available, does not validate their use in the 2008 Smelt BiOp.

d.

FWS’s Comparison of CALSIM II Data to DAYFLOW Data.

The BiOp’s effects analysis used analytical methods and data, “including the CALSIM II model outputs provided in the appendices of Reclamation’s 2008 OCAP BA, historical hydrologic data provided in the DAYFLOW database, statistical summaries derived from 936 unique 90-day particle tracking simulations published by

*897

Kimmerer and Nobriga (2008), and statistical summaries and derivative analyses of hydrodynamic and fisheries data published by Feyrer et al. (2007), Kimmerer (2008), and Grimaldo et al. (accepted manuscript).” BiOp at 204.

CalSim II is a computer model developed jointly by DWR and Reclamation. Declaration of Aaron Miller,

21

Doc. 548-1, at ¶ 5. The model simulates SWP and CVP operations and is the standard planning tool for evaluating project operations.

Id.

at ¶ 6. CalSim II has been continuously updated since it was first applied in 2002.

Id. at

¶ 8. CalSim II simulates SWP and CVP reservoir operations, project exports and water deliveries, flow through the Delta, and salinity requirements in the Delta, including the location of X2.

Id.

at ¶ 7.

CalSim II uses historic hydrologic data from October 1922 to September 2003, including precipitation, runoff into reservoirs and inflow into the Delta from unimpaired streams. Miller Decl., Doc. 548-1, at ¶ 10 & n. 1. The model further assumes a level of development, which reflects water demand resulting from particular levels of urban population, agricultural production, and wildlife refuge needs,

id.

at ¶ 10, along with the effect of environmental regulations and programs,

id.

at ¶ 27; BiOp at 207. CalSim II is capable of estimating the position of X2. Miller Decl., Doc. 548-1, at ¶ 14.

The BiOp considered a number of CalSim II studies, either directly or indirectly:

• Study 6.0 was designed to represent the assumptions used in the 2004 OCAP BA within the updated CalSim II model framework in order to highlight changes from the previous model framework. This Study models a 2005 level of development and includes steps to account for operations under CVPIA (b)(2) and Joint Point of Diversion

22

.

See

OCAP BA at 9-32 (AR 010729).

• Study 6.1 is similar to 6.0, except that the 2005 Trinty River Record of Decision is removed, and the Joint Point of Diversion is not accounted for.

Id.

• Study 7.0 was developed as the baseline study for the OCAP BA. Study 7.0 represents existing conditions, and assumes a 2005 level of development and a full environmental water account (“EWA”)

23

. BiOp at 207.

• Study 7.1 is a near-future conditions study. It assumes a 2005 level of development and a limited EWA. BiOp at 207-08.

• Study 8.0 is a future conditions study. It assumes a 2030 level of development and a limited EWA. BiOp at 208.

*898

• The 9.0 series of studies represents climate change scenarios. BiOp at 208.

The OCAP BA suggested using Calsim II Study 7.0 as the current baseline and Study 6.1 as the historical baseline for evaluating the impacts of project operations. BiOp at 204. However, the BiOp rejected this suggestion because, although “changes were expected between Study 6.1 and Studies 7.0 and 7.1,” the modeled results were “nearly identical.”

Id.

FWS concluded from this result that Calsim II could not accurately generate an empirical baseline.

See id.

at 204-06. Instead, FWS chose to “use actual data to develop an empirical baseline,” including the use of the Dayflow model to “develop[ ] historical time series data for hydrologic variables.” BiOp at 206. Dayflow is a model that estimates historic outflow based on historic precipitation, inflow, and exports, and estimates of delta island diversions. Dayflow also provides an estimate for the location of X2. Miller Decl., Doc. 548-1, at ¶¶ 14-15.

In the BiOp, FWS purports to quantify adult entrainment by comparing OMR flows from CalSim II studies to historic OMR flows during 1967-2007. BiOp at 212-13. The BiOp depicts these results in Tables E-5a, E-5b, and E-5e:

*899

[[Image here]]

Tables E-5b and E-5c depict changes in OMR flows and entrainment using the Dayflow-generated historic data as the baseline and comparing that to CalSim II study results. In addition, the BiOp utilized an equation taken from Kimmerer’s

*900

2008 paper to estimate the population loss of delta smelt under the various modeled scenarios. The results of these ealculations were depicted in Figures E-5 and E-6:

Fagace 5-5

Fretjtunty

dtitifttaiem of predicted adult debs -imtílí entwined a? 8?g*:s asid Jane: for predicted «¡íimams ác-mlmtcsjc fa <¶9$7-1994;« actual estimate: Sctu KtoUtiíuyri.ZCOSl dar jean 1935-211015. and those «timriHS. ftomCALS&lIE mode! data bv StvJv.

[[Image here]]

Figure E-á. Same as E-5 but by water year type. Krmmerer (2008) estimates did not include below normal or critical dry water year types.

[[Image here]]

*901

BiOp at 251-52. The accompanying text explains the significance of these results to the analysis:

The median OMR flows from the CALSIM II modeled scenarios were more negative than historic OMR flow for all WY types except critically dry years (Figure E-3; see Table E-5b for all differences). Overall, proposed OMR flows are likely to generate increases in population losses compared to historic years (Figure E-5 and Figure E-6). For example, the frequency of years when population losses are less than 10 percent from most modeled studies (except studies 7.0 and 8.0) is less than 24 percent compared to historic estimates that only exceed 10 percent in approximately half of the years.

The most pronounced differences occur during wet years, where median OMR flows are projected to be approximately 400 to 600 percent (-7100 to -3678 cfs) higher than historical wet years (-1032 cfs). Generally, wet years are marked by low salvage and population losses. However, the proposed operations during wet year are predicted to cause up to a 65 percent increase in smelt salvage and lower probability that population losses will be below 10 percent.

The proposed operation conditions likely to have the greatest impact on delta smelt are those modeled during above normal WYs. The modeled OMR flows for the above normal WYs ranged between -8155 and -6242 cfs, a 33 to 57 percent decrease from the historic median of 5178 cfs. Though the predicted salvage would only be about 15-20 percent higher than historic salvage during these years (Table E-5e), the modeled OMR flows in these years would increase population losses compared to historic years.

In below normal and dry WYs, proposed OMR flows are also modeled to decrease from historic medians. Predicted salvage levels are likely to increase between 2 and 44 percent. More importantly, the modeled median flows from all studies in these WY types range between -5747 and -7438 cfs. Modeled OMR flows at these levels are predicted to increase salvage and increase the population losses from historic levels as well.

During critically dry years, the median OMR flows for studies 7.0, 7.1, 8.0, 9.1, 9.4, and 9.5 are less than -5,000 cfs. These studies have predicted salvage lower than historic salvage and are not likely to generate larger population losses compared to historic years. The models might overestimate salvage during critical dry years when smelt are unlikely to migrate towards the Central Delta due to lack of turbidity or first flush. Thus, the effects of critical dry operations on delta smelt take are probably small and lower than estimated.

BiOp at 212-13.

Based on these comparisons of CalSim II data and Dayflow-generated historic data, the BiOp concludes, “adult entrainment is likely to be higher than it has been in the past under most operating scenarios, resulting in lower potential production of early life history stages in the spring in some years.” BiOp at 213.

The BiOp performed comparisons of CalSim II data to Dayflow-simulated historic baseline data to quantify the effects of the action on larval and juvenile delta smelt.

See, e.g.,

BiOp at 219 (examining effect of action on larval and juvenile entrainment: “[t]he analysis is based on comparison of historical (1967-2007) OMR and X2 to the proposed action’s predictions of these variables provided in ... [CalSim] studies 7.0, 7.1, 8.0, and 9.0-9.5”). Figure

*902

E-18 depicts several sets of calculations of the frequency at which certain percentages of the delta smelt population would be entrained:

figsie £-17. Frequency d:7trifcunz?u ofeartmafedpropc-'nic-iir

c-fm

smelt entsained at

hml'A mú

Jo&ei &r ¿PfrT-í 994

mé

K‘?5-N1&7. Tie dan were

exmvohzm to

an 37-Tear

mnoá is

make újentcontoaiabie :s the CAUSE*! 2 out

[[Image here]]

BiOp at 264. The black dashed line depicts entrainment estimates for Dayflowgenerated historic data from 1967 to 1994, the red line with diamonds depicts entrainment estimates for Dayflow-generated historic data from 1995-2007, and the fine lines depict the various entrainment estimates based on Calsim II data. Based on these calculations, the BiOp concludes that “the proposed action will decrease the frequency of years in which estimated entrainment is < 15 percent. Thus, over a given span of years, the project as proposed will increase larval juvenile entrainment relative to 1995-2005 levels. This will have an adverse effect on delta smelt based on their current low population levels.” BiOp at 222.

A separate BiOp analysis purports to quantify the effects of the project operations on delta smelt

habitat

by comparing CalSim II model projections of the location of X2 under the proposed operations to the median location of X2 over the historical period 1967-2007, as simulated by Day-flow. BiOp at 235-36. Based on this comparison, the BiOp concludes “[t]he median X2 [locations] across the CalSim II modeled scenarios were 10-15 percent further upstream than actual historic X2 (Figure E-19).”

Id.

at 235. In reliance on these percent differences between CalSim II-created data and historical data, the BiOp concludes: “proposed action operations are likely to negatively affect the abundance of delta smelt.”

Id.

at 236.

According to Plaintiffs, the comparison of Calsim II to Dayflow outputs distorts the BiOp in several key ways:

(1) The comparison of outputs of these two models in the Project Effects analysis is,

ipso facto,

a violation of the best available science requirement.

*903

(2) To use Dayflow, which represents historical conditions, to generate the baseline for the Project Effects analysis, improperly attributes past effects to the Projects;

(3) Because the flawed comparison was used to support imposition of Component 3 (Action 4) (a/k/a the “fall X2” action), that Action is invalid.

24

(1)

Was FWS’s Decision to Compare Calsim II to Dayflow Model Runs a Violation of the Best Available Science Requirement?

Mr. Aaron Miller opines that outputs from a CalSim II study should not be compared to outputs from the Dayflow model because the assumptions used in the two models are significantly different. Miller Deck, Doc. 548-1, at ¶¶ 22-55. He identified the following key differences between the models:

• Level of Development: The CalSim II model assumes a constant level of development. In contrast, the Dayflow model incorporates a continuous change in the level of development because the Dayflow model is using historical information as input. "When comparing models to determine the effect of project operations, the best scientific practice is to keep the assumed level of development constant.

Id.

at ¶¶ 31-38.

• Regulatory Assumptions: CalSim II assumes a constant regulatory environment, whereas Dayflow uses a regulatory environment that has changed over time. Over the past 40 years, numerous regulatory programs have altered the way the projects are operated, including D-1485, D-1641, the Central Valley Project Improvement Act (“CVPIA”), the 1995 Water Quality Control Plan, and the EWA. These differences “further undermine the reliability of comparing historically based Dayflow values to the Calsim II model results.”

Id.

at ¶¶ 39-41.

• Time Step: CalSim II operates on a monthly time step, whereas Dayflow operates on a daily time step.

Id.

at ¶ 42.

• Operational/Computational Guidelines: The Dayflow model incorporates real-world conservative operational tactics designed to avoid violating applicable regulations. In contrast, the CalSim II model operates strictly to that regulation.

Id.

at ¶ 44. Operating conservatively results in higher modeled outflow.

Id.

• Year Range: The Dayflow model uses a different historic time window than CALSIM II. The BiOp used values from 1967 to 2007 as inputs into the Dayflow model, while 1922 to 2003 were used for Calsim II.

Id.

at ¶ 52. This introduces additional error into any comparison between outputs of these two models because the time period used for the Dayflow model had a higher percentage of wet or above normal years, as compared to the time period covered by Calsim II.

Id.

at ¶ 53.

• Method for Calculating position of X2: The artificial neural network

*904

(“ANN”)and the Kimmerer Monismith equation (“KM equation”) are two methods of estimating X2.

Id.

at ¶ 46. The CalSim II studies used ANN to estimate the position of X2, while the Dayflow model uses the KM equation.

Id.

at ¶ 47. Holding all other variables constant, but varying the method

(ANN v. KM)

used, produces inconsistent results. At locations less than 75 kilometers (“km”)from the Golden Gate, the KM equation results in an X2 estimate greater than (or farther upstream than) the ANN estimate. In contrast, at locations greater than 75 km from the Golden Gate, the KM equation provides an estimate less than the ANN estimate.

Id.

at 11, Fig. 2.

Mr. Miller opined that best scientific practice is to compare models that use consistent assumptions and methodologies.

See id.

at ¶¶ 38, 51, 54;

see also id.

at ¶ 41. The approach taken in the BiOp, quantitatively comparing Calsim II runs to Day-flow model outputs “introduces significant error into the analysis.”

Id.

at ¶ 56.

Dr. Punt, a 706 Expert added that “[i]n principle, there is nothing wrong with fitting a model using a set of OMR/X2 valued from one model and making predictions using OMR/X2 values which are based on the output from a different model, as long as the two sets of values are calibrated.... Not calibrating the two sets of model outputs will lead to some bias in the inferences, with the level of bias dependent on the net effect of all the differences between the ‘historical’ and Calsim II values for the same years.” Doc. 633-1 at 15.

Mr. Derek Hilts, a FWS employee who previously served as “Engineer-in-Charge” of CVP/SWP modeling for Reclamation, disagrees with Mr. Miller’s general opinion that comparing Calsim II and Day-flow outputs is

per se

scientifically unreliable, noting that the OCAP BA’s Appendix D specifically compared Calsim II and Dayflow runs for the purposes of testing “Calsim II’s ability to simulate the CVP/ SWP system reasonably well.” Deck of Derek Hilts, Doc. 540, at ¶ 11. But, as Mr. Miller explains, this type of “validation comparison” is designed to “help establish the credibility of the CalSim II model by showing that the model moves water, simulates operation of the export pumps, and so forth, with the same general timing and magnitude as actual historical data show.” Second Miller Deck, Doc. 597, at ¶ 12. In fact, Mr. Miller points out that the detailed validation data contained in the OCAP BA demonstrate that, although Calsim II outputs generally track historical data, they “do not precisely match the actual historical data.”

Id.

at ¶ 12. Because validation is “looking only at the general operational performance of the model,”' a validation comparison “does not need to control for the effects of all the differences in the model and the historical measurements

....” Id.

at ¶ 13.

More specifically, Mr. Hilts disagrees with Mr. Miller’s critique that the divergent methods of calculating the position of X2 render the comparison used in the BiOp scientifically inappropriate. Mr. Hilts does not dispute Mr. Miller’s conclusion that the KM and ANN equations produce marginally different outcomes. Instead, Mr. Hilts criticizes Mr. Miller for failing to “assert that any such error would have changed the conclusions drawn in the BiOp.” Doc. 540 at ¶ 19.

Assumedly to demonstrate that the conclusion would not have changed, Mr. Hilts revisited the calculations in the BiOp, using the KM equation in

both

models to

*905

produce revised estimates of the position of X2.

25

In performing this analysis, Mr. Hilts also attempted to correct for one of the other purported sources of bias-the inconsistent year range—as well as for a few incorrect data points found in the underlying data used in the BiOp. Doc. 540 at ¶¶ 17-18. This revised analysis, which is presented in Exhibit 2, Figure 2 to Mr. Hilts’ declaration, is replicated below:

[[Image here]]

Doc. 540, Exhibit 2, Figure 2. According to Mr. Hilts, this figure demonstrates the “same general upstream movement” of X2 “discussed in the 2008 BiOp.”

Id.

at ¶ 17.

26

Recognizing that his revised analysis demonstrates the same general upstream shift as the BiOp, Mr. Hilts criticizes Mr. Miller for failing to “quantify the effect of the alleged biases ostensibly embedded in the X2 comparison presented in the BiOp.”

Id.

at ¶ 7. Federal Defendants contend that even if the Calsim II to Dayflow comparison introduced bias, that bias was not significant. However, the record suggests otherwise.

Recognizing that it is not possible to quantify all aspects of the error caused by the comparison of Calsim II runs to Day-flow output, Mr. Miller’s reply declaration endeavored to quantify the bias in his reply declaration.

See

Second Miller Decl., Doc. 597. As with Mr. Hilts’ revised calculations, Mr. Miller compared the results reported in the BiOp (Calsim II runs applying the ANN equation and Dayflow runs using the KM equation), to a revised set of results using the KM equation in

*906

stead of ANN in the Calsim II runs.

Id.

at ¶ 14. Mr. Miller’s analysis shows that project operations will cause an upstream shift in X2. Mr. Miller explained that the BiOp’s comparison reflected a difference between the reported historic median of X2 [79 km] and the study 7.0 median [87 km] of 10% [ (87 km-79 km)/79]. Mr. Miller concluded that the median X2 for the CalSim 7.0 study using the KM equation (instead of using ANN) was 84 km (instead of 87 km). Finally, he identified the percent difference between the reported historic median estimate of X2 using the KM equation [79 km] and the CalSim study 7.0 median estimate of X2 using the KM equation [84 km] to be 6% [ (84 km-79 km)/79 km].

Id.

at ¶ 14; BiOp at 235-36. From this, Mr. Miller concluded 40% of the difference between X2 as estimated by study 7.0 and the historical X2 baseline reported in the BiOp is error attributed entirely to the use of the KM equation to calculate the historical baseline X2 and the ANN equation to calculate the CalSim II study 7.0 results.

Id.

at ¶ 15. It is unknown which portion of the remaining 60% of difference is attributable to the proposed action, and which portion is due to the other identified biases.

Id.

at ¶ 16. Dr. Punt expressed a corroborating opinion, estimating that the bias created by failing to calibrate the models “seems non-trivial” and opining that it could be “as large as the differences seen in Figure E-19,” the figure in the BiOp depicting the purported 10% shift in X2 between the historic/Dayflow runs and the Calsim II runs. Doc. 633-1 at 16. Following a similar methodology, using the BiOp’s Figure E-20 equation, Mr. Miller calculated the reduction in suitable habitat consistent with the change in the position of X2. A comparison of CalSim II study 7.0 with study 7.1 yielded a reduction in habitat area of 128 hectares (or 2.8%), and a comparison of study 7.0 with study 8.0 yielded a reduction in habitat area of 289 hectares (or 6.2%). Doc. 597 at ¶ 20; BiOp at 266.

Mr. Miller opined that all errors/biases could have been avoided by comparing CalSim II study 7.0—designed as a current conditions baseline—instead of the “historical” baseline in the BiOp, to the near-future 7.1 study.

27

However, Mr. Hilts points out that comparing Calsim II Study 7.0 to 7.1 and 8.0 is simply “not responsive to the need for comparisons with historical X2 locations,” because none of the Calsim II simulations represent Delta conditions that existed from 1967-2007. Doc. 540 at ¶ 9. “With the Fall X2 comparison, [ ]FWS wanted to investigate whether the continuation of the recent, as well as future, CVP/SWP operations would result in less or deteriorated habitat for delta smelt relative to the habitat that prevailed historically.”

Id.

at ¶ 8. “The CalSim II simulations that Mr. Miller would have the FWS use do not” accomplish this.

Id.

The theoretical problems with using a Calsim II to Calsim II comparison were manifest. As discussed above, when CalSim II was used to model current Project operations, and these results were then compared to the results of a CalSim II modeling run purportedly simulating past operations, the results “were nearly identical” despite significant operational changes in current operations as compared to past. BiOp at 204-205. The BiOp explains that

*907

“[t]he inaccuracies in CalSim [led FWS] to use actual data to develop an empirical baseline.”

Id.

at 206.

28

FWS contends it had legitimate reasons to rely on a Calsim II to Dayflow comparison instead of a Calsim II to Calsim II comparison.

In light of the known and material resulting disparity, FWS’s decision to use a Calsim II to Dayflow comparison to

quantitatively

justify its jeopardy and adverse modification conclusions, without attempting to calibrate the two models or otherwise address the bias created, was arbitrary and capricious and ignored the best available science showing that a bias was present. The BiOp specifically relied upon the

quantitative

nature of the Calsim II to Dayflow comparisons in many places. For example, in reference to the X2 shift and resulting effects on smelt habitat:

The median X2 across the CALSIM II modeled scenarios were 10-15 percent further upstream than actual historic X2 (Figure E-19). Median historic fall X2 was 79km, while median values for the CALSIM II modeled scenarios ranged from 87 to 91km. The CALSIM II modeled scenarios all had an upper range of X2 at about 90km. The consistent upper cap on X2 shows that water quality requirements for the Delta ultimately constrain the upper limit of X2 in the simulations. These results were also consistent across WY types (Figure E-19) with the differences becoming much more pronounced as years became drier. Thus, the proposed action operations will affect X2 by shifting it upstream in all years, and the effect is exacerbated in drier years.

BiOp at 235. The BiOp does not explain to what extent the ultimate jeopardy/adverse modification conclusions were based upon the calculated magnitude (10-15 percent) of the X2 shift, rather than the existence of a shift. It cannot be determined whether the BiOp would have reached the same conclusion had this bias not been present.

Federal Defendants concede but understate that “the two models are not perfectly calibrated, and a slight transformation of the data occurs when the analysis switches from one model to the other, the BiOp acknowledges this slight shift.” Doc. 660 at 36. Nevertheless, FWS concluded in its “scientific judgment [ ] that the CalSim [II]-to-Calsim [II] output was far worse.”

Id.

(citing BiOp at 207). Federal Defendants argue this was a choice between “one comparison that yielded a slight calibration issue and another that completely masked altogether the variable sought to be compared .... ” and that “it would have been irrational for the Service to proceed with [a Calsim II to Calsim II comparison” after discovering its flaws.

Id.

This may be the case, but it does not follow that what FWS did with the Calsim II to Dayflow comparisons was rational or based upon the best available science.

FWS had actual notice of scientific concerns with comparing historical data to CalSim II simulated data. DWR Deputy Director Jerry Johns, on October 24, 2008, submitted comments to FWS on the draft effects analysis, generally cautioning against the comparison of modeled data with actual data:

USFWS is using historic data for comparison to CalSim II simulations. Great

*908

caution should be taken when comparing actual data to modeled data. CalSim II modeling should be used in a comparative mode. In other words, it should be used to compare one set of model runs to another. For example, it would be appropriate to compare CalSim II modeling of one demand alternative to another to analyze the incremental effects.

AR 008671;

see also

AR 008668 (further explaining unreliability problems comparing historic and modeled data). Although neither Mr. Miller nor any interested party suggested that comparing Dayflow to Calsim II data was a scientifically invalid methodology prior to the issuance of the BiOp, the BiOp does not recognize the essential methodological defect, or explain how any of the conclusions it reached account for it. Nor does the BiOp explain how it is able to attribute the changes in X2 it found between the “historic” baseline and the CALSIM studies to the proposed action, and not to any of the other differences between the Dayflow and Calsim II models. Instead, FWS only rationalizes that it opted to use the “historic” baseline rather than CALSIM Study 7.0 as the baseline because, “the CALSIM monthly simulation model does not capture a precise Delta operation.... [Thus], the inaccuracies in CALSIM lead us to use actual data to develop an empirical baseline.” BiOp at 204 & 206. This statement may explain the reasons for FWS’s decision, but it does not justify its ultimate conclusion.

This is of particular concern because DWR, a joint operator of the projects communicated its scientific and operational concerns based on known available science. DWR and Reclamation have legal obligations to allocate water supply reasonably and responsibly, not solely to save the species. As discussed in below at Part VII.B, FWS’s focus on its responsibilities to the species appears to have caused it to ignore its own regulations’ obligations to consider impacts to the overall water supply and additional uses. The potential impacts of inaccurate quantitative analyses in the BiOp cannot be understated.

Defendants argue FWS’s decision to compare the two models to quantify the shift of X2 was a reasonable scientific decision, even though other experts may disagree. Doc. 660 at 17-19; Doc. 661-3 at 13-14. Federal Defendants cite

Lands Council,

537 F.3d at 993, to justify FWS’s modeling decisions as entitled to deference, because it is a matter “within its area of special expertise, at the frontiers of science.”

29

As a general rule, choices regarding modeling methods are exactly the sort of choices that, under the APA, are left to the expert agency in the exercise of its discretion.

NWF v. EPA,

286 F.3d at 565 . A court “may reject an agency’s choice of a scientific model only when the model bears no rational relationship to the characteristics of the data to which it is applied.”

Id.

at 565 (internal quotations and citations omitted).

Lands Council

in

*909

structs that a court is “not free to impose on the agency [its] own notion of which procedures are best.... Nor may [it] impose procedural requirements not explicitly enumerated in the pertinent statutes.” 537 F.3d at 993 (internal citations and quotations omitted);

id.

at 1000 (finding agency did not act arbitrarily “in relying on its own data and discounting the alternative evidence offered” by plaintiffs because “[w]hen specialists express conflicting views, an agency must have discretion to rely on the reasonable opinions of its own qualified experts even if, as an original matter, a court might find contrary views more persuasive”) (citations omitted).

In

NWF v. EPA,

the EPA evaluated several regulatory options for economic feasibility, applying a particular model to predict whether businesses were likely to go bankrupt under the weight of additional regulation. NWF criticized the model on several grounds, including that the model had “an error rate of at least 15%.”

Id.

at 565. The D.C. Circuit examined and rejected each critique, reasoning that none called into question the model’s reliability.

Id.

Here, however, undisputed expert testimony offered by DWR, a co-operator of the Projects, calls into question the manner by which FWS utilized the two models to evaluate the impact of project operations on the position of X2. The Calsim II model was developed by DWR and Reclamation as a planning tool to simulate State Water Project and Central Valley Project operations. DWR, one of the agencies with special expertise in the use and application of Calsim II,

see

BiOp at 207; Miller Decl., Doc. 548-1, at ¶ 5-7, raised cautions and objects to the manner in which FWS used the model. Federal Defendants do not rebut the undisputed expert evidence that using such comparisons for quantitative purposes is scientifically improper. All experts in this case agree that data from two different models should not be compared without calibration. Doc. 633-1 at 13-17 (706 expert report); Miller Decl., Doc. 548-1, ¶¶ 22-55; Second Miller Decl., Doc. 597, ¶¶ 4-22. In other words, even though no superior set of models have been identified, the chosen models were indiscriminately used without addressing an important factor, the potential (and apparently real and significant) bias created when the results of two different computer models were used to perform quantitative comparisons. Unlike

NWF v.

EPA, where the agency applied a model that was deemed reliable, here, FWS has not addressed or explained the material bias created by its methodological choices. It cannot be determined whether FWS would have reached the same result had the bias been considered or addressed. FWS must do so on remand.

(2)

Does the Use of Day flow to Represent the Baseline in the Project Effects Analysis Improperly Attribute Past Effects to the Projects?

DWR asserts that FWS’s use of an “historical baseline” was per se unlawful because the ESA’s implementing regulations “require the Service to use current operations, not past operations, as the baseline for its effects analysis.” Doc. 548 at 7-8. In support of this contention, DWR cites 50 C.F.R. § 402.02 , which defines the “environmental baseline” to include:

the past and present impacts of all Federal, State, or private actions and other human activities in the action area, the anticipated impacts of all proposed Federal projects in the action area that have already undergone formal or early sec

*910

tion 7 consultation, and the impact of State or private actions which are contemporaneous with the consultation in process.

See also

Consultation Handbook at 4-22 (baseline includes “effects of past and

ongoing

human and natural factors leading to the current status of the species”) (emphasis added). In addition, DWR cites

NWF v. NMFS II,

524 F.3d at 930 , which held that an agency action “only ‘jeopardized’ a species if it causes some

new

jeopardy.” (Emphasis added.) DWR argues that “[bjecause [FWS’s] baseline looks to decades past, it cannot be used as a basis for assessing any ‘new jeopardy’ posed by Project operations going forward.” Doc. 548 at 8.

30

DWR oversimplifies the issue. FWS’s BiOp sought to determine whether

ongoing

and future coordinated operations of the CVP and SWP would cause jeopardy to the delta smelt or adversely affect its critical habitat. Arbitrarily setting the baseline at 2008, when the BiOp’s analysis was finalized, would not have captured the impacts of then-ongoing project operations. The agency had discretion to use a historic baseline.

(3)

Use of Comparisons Between CALSIM and DAYFLOW Model Outputs to Justify Imposition of Component 3 (Action I), the Fall X2 Action.

In addition to utilizing comparisons of Calsim II and Dayflow data in the Project Effects section to demonstrate that Project Operations affect the location of X2, the BiOp relies on these comparisons to justify the imposition of RPA Component 3 (Action 4, or the “Fall X2 action”). The BiOp’s “Justification” section discussing Action 4 references the Calsim II to Day-flow comparison:

The Effects section clearly indicates there will be significant adverse impacts on X2, which is a surrogate indicator of habitat suitability and availability for delta smelt in all years (Figures E-19 and E-25 in Effects section).... The action is focused on wet and above normal years because these are the years in which project operations have most significantly adversely affected fall (Figure E-27 in Effects section) and therefore, actions in these years are more likely to benefit delta smelt.

BiOp at 373. Figures E-19 and E-25 compare historic X2 locations simulated by Dayflow to conditions under planned project operations simulated by Calsim II:

*911

Fjgns* 34°

XI ikm)

cHuie ro Dfe?mb$r Kawr"

ca iamr.c fan má

CAI5JN! H ur-M j«7jiK Tlie •;*;rr?r line i»r> tas ii* median

n~.ñ

4j?

.y,mi

Xn l'Oi-.iíáiiw',,«« it? fir,’- iiuc ^aarHis*;

[[Image here]]

*912

Figure E-25. Smoothed trend lines for the time series of historic and CALSIM II-modeled fall N2.

[[Image here]]

BiOp at 265, 271.

Undisputed expert testimony establishes the likelihood that the comparison of Day-flow to Calsim II data introduced significant error into the analysis that forms the basis for Figures E-19 and E-25. Mr. Miller concluded 40% of the difference between X2 as estimated by study 7.0 and the historical X2 baseline reported in the BiOp is error attributed entirely to the use of the KM equation to calculate the historical baseline X2 and the ANN equation to calculate the CalSim II study 7.0 results. Second Miller Decl., Doc. 597, at ¶ 15. It is unknown which portion of the remaining 60% of difference is attributable to the proposed action, and which portion is due to the other identified biases.

Id.

at ¶ 16. Dr. Punt gave a consistent opinion, estimating that the bias created by failing to calibrate the models “seems non-trivial” and opining that it could be “as large as the differences seen in Figure E-19,” the figure in the BiOp depicting the shift in X2 between the historic/Dayflow runs and the Calsim II runs. Doc. 633-1 at 16.

Federal Defendants do not respond directly to these assertions of bias. Instead, they point out that the historical X2 data was not the only basis for Action 4. Doc. 660 at 49. The BiOp describes multiple sources of information that were considered:

This analysis of the effects [of the] proposed CVP and SWP operations on the delta smelt and its critical habitat uses a combination of available tools and data, including the CALSIM II model outputs provided in the appendices of Reclamation’s 2008 Biological Assessment, historical hydrologic data provided in the DAYFLOW database, statistical summaries derived from 936 unique 90-day particle tracking simulations published by Kimmerer and Nobriga (2008), and statistical summaries and derivative analyses of hydrodynamic and fisheries data provided by Feyrer et al. (2007), Kimmerer (2008), and Grimaldo, et al. (accepted manuscript).

BiOp at 204;

see also

Feyrer Decl., Doc. 541, at ¶ 17. Additionally, “[t]he Service’s

*913

examination of habitat suitability during fall is derived from published literature and unpublished information linking X2 to the amount of suitable abiotic habitat for delta smelt (Feyrer et al. 2007, 2008).” BiOp at 234. The BiOp expressly recognizes that the modeling does not precisely represent historic X2, as do the peer-reviewed studies on which the BiOp relies in part for this component.

See

BiOp at 204; AR 018278-018306 (Feyrer,

et al.

(2008)).

The justification for Action 4 relies heavily on the quantitative analyses presented in Figures E-19 and E-25.

See

BiOp at 373. Whether Action 4, which has substantial adverse impacts on the water supply, is justified in the absence of the quantitative analysis cannot be determined. These questions are too serious to go unanswered and must be remanded to the agency for further explanation and/or correction.

(3)

Other Challenges to the Fall X2 Action.

Plaintiffs raise additional challenges to the justification for the Fall X2 action, arguing “neither the BiOp nor the record demonstrate that Component 3 (Action 4) is necessary to avoid jeopardy to the delta smelt or destruction or adverse modification of its critical habitat, or that it will materially benefit the species or its habitat.” Doc. 697 at 25.

a.

Plaintiffs’ Argument that Action k is an “Untested Hypothesis. ”

Plaintiffs maintain that Action 4 is nothing more than an “untested hypothesis,” emphasizing that FWS acknowledges the need to assess the efficacy of Action 4 over time:

The Service shall conduct a comprehensive review of the outcomes of the Action and the effectiveness of the adaptive management program ten years from the signing of the biological opinion, or sooner if circumstances warrant. This review shall entail an independent peer review of the Action. The purposes of the review shall be to evaluate the overall benefits of the Action and to evaluate the effectiveness of the adaptive management program. At the end of 10 years or sooner, this action, based on the peer review and Service determination as to its efficacy shall either be continued, modified or terminated.

BiOp at 283.

This does not render Action 3 a mere “hypothesis,” nor does this “demonstrate] the absence of a rational connection between Action 4 and an increase in smelt abundance.” Doc. 697 at 25. It is not inconsistent to find an action necessary, while also calling for an evaluation whether that action actually produced the expected outcomes. It is of no moment that in a research paper Mr. Feyrer referred to the X2 requirement as “the hypothesis that the combined effects of pre-adult abundance and the amount of suitable abiotic habitat (or X2) during autumn affect recruit abundance the following summer.” AR 018285 (Feyrer unpub. 2008). He is a scientist gathering further information about the relationship between X2 and smelt population dynamics. The record does not suggest this is scientifically improper. It was not clearly erroneous for FWS to rely upon Feyrer’s 2008 research paper.

b.

FWS’ Reliance on the Feyrer Papers.

FWS based its effects analysis of X2 in part

31

on two articles written by Feyrer et

*914

al., which purported to show a correlation between X2 in the autumn and subsequent delta smelt abundance.

See

BiOp at 235-38 (citing Feyrer et al. (2007); Feyrer et al. (2008)). Plaintiffs argue that these articles did not represent the best available science because “the correlation they claimed to find was driven by the presence of a single unrepresentative data point.” Doc. 551 at 34. Even assuming the scientific validity of the 2007 and 2008 Feyrer analysis, Plaintiffs contend the BiOp’s X2 conclusions far exceed what the articles scientifically support.

Id.

Plaintiffs’ letter, responding to a draft of the BiOp, identified a purported flaw in the Feyrer et al. (2008) analysis: the supposed correlation between Fall X2 and delta smelt abundance Feyrer et al. was driven by the presence of a single, apparently outlier, data point. Removing that data point resulted in a finding of no statistically significant relationship between Fall X2 and the abundance of delta smelt.

See

SLDMWA & SWC Letter to NMFS and FWS (Oct. 20, 2008) at 2 (AR 006407). As the letter noted, “a correlation solely reliant upon a single data point cannot reasonably be considered as an actual indicator of cause.”

Id.

Plaintiffs’ argument continues:

That there was no statistically significant relationship between X2 and delta smelt abundance during the 1987-2007 period should not have been surprising given that Feyrer et al. found no statistically significant relationship between the two factors for the 1968-1986 period or for the entire 1968-2007 period. Feyrer et al. (2008) at 14 (AR 018291). Nor was it surprising considering that— as the Feyrer et al. (2008) article conceded—the existing best available science on delta smelt showed no direct correlation between the location of Fall X2 and delta smelt abundance. Feyrer et al. (2008) at 8 (“[P]revious analyses have not shown simple relationships between X2 and delta smelt abundance.”) (AR 018285).

Doc. 551 at 35.

Federal Defendants respond:

[Ujnless data points are excluded to control for a specific variable, or for some other explicit reason that is central to measuring the relationship at issue,

*915

there is no scientific reason to remove a data point from an analysis just because it changes the result. In any event, removing the data point challenged by Plaintiffs does not appreciably change the result—the result goes from a 95% probability the relationship is not due to chance to a 92% probability that the relationship is not due to chance. Moreover, this is an argument that can go both ways. Removing other individual data points would increase the statistical significance.

Doc. 660 at 44. Federal Defendants are correct that removing a data point simply because it changes the result would be arbitrary. Plaintiffs do not point to any scientific basis, let alone an undisputed one, for excluding the so-called “outlier” point, other than that it is an outlier. Plaintiffs do not show the point is erroneous or identify competing studies that reach different opinions from Feyrer that FWS failed to consider. This is a scientific dispute among experts over which the agency is owed deference.

c.

Do the Studies Cited in the BiOp Support FWS’s Conclusion that Fall X2 Determines the Extent of Suitable Smelt Habitat?

The BiOp concludes that to avoid jeopardy the RPA Actions must “[ijmprove fall habitat for delta smelt by managing [ ] X2 through increasing Delta outflow during fall when the preceding water year was wetter than normal.” BiOp at 369;

see also

BiOp at 374 (“Outflow during fall determines the location of X2, which determines the amount of suitable abiotic habitat available to delta smelt (Feyrer et al. 2007, 2008).”). Plaintiffs argue that none of the articles FWS cited in the BiOp actually support FWS’s conclusion that the location of X2 determines the amount of suitable habitat for the delta smelt.

See

Doc. 551 at 39-41.

(1)

Feyrer (2007).

Plaintiffs first criticize the BiOp’s reliance on a 2007 Canadian Journal of Fisheries and Aquatic Sciences paper by Feyrer, Nobriga, and Sommer, three scientists then working for Plaintiff DWR, entitled, “Multidecadal trends for three declining fish species: habitat patterns and mechanisms in the San Franci

This text is long and has been trimmed here. Open the source document for the complete record.